commit 53d79be8bd9b7b06d2442b08a5c3e22adb6ea810 Author: ModelHub XC Date: Wed Jun 3 17:10:12 2026 +0800 初始化项目,由ModelHub XC社区提供模型 Model: laion/openthoughts-4-code-qwen3-32b-annotated-7k_qwen3-1.7B_10k Source: Original Platform diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..3d5b1c8 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,53 @@ +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text + + +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ftz filter=lfs diff=lfs merge=lfs -text +*.gz filter=lfs diff=lfs merge=lfs -text +*.h5 filter=lfs diff=lfs merge=lfs -text +*.joblib filter=lfs diff=lfs merge=lfs -text +*.lfs.* filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.onnx filter=lfs diff=lfs merge=lfs -text +*.ot filter=lfs diff=lfs merge=lfs -text +*.parquet filter=lfs diff=lfs merge=lfs -text +*.pb filter=lfs diff=lfs merge=lfs -text +*.pt filter=lfs diff=lfs merge=lfs -text +*.pth filter=lfs diff=lfs merge=lfs -text +*.rar filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* filter=lfs diff=lfs merge=lfs -text +*.tflite filter=lfs diff=lfs merge=lfs -text +*.tgz filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip filter=lfs diff=lfs merge=lfs -text +*.zstandard filter=lfs diff=lfs merge=lfs -text +*.tfevents* filter=lfs diff=lfs merge=lfs -text +*.db* filter=lfs diff=lfs merge=lfs -text +*.ark* filter=lfs diff=lfs merge=lfs -text +**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text +**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text +**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text + +*.ckpt filter=lfs diff=lfs merge=lfs -text +*.gguf* filter=lfs diff=lfs merge=lfs -text +*.ggml filter=lfs diff=lfs merge=lfs -text +*.llamafile* filter=lfs diff=lfs merge=lfs -text +*.pt2 filter=lfs diff=lfs merge=lfs -text +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz filter=lfs diff=lfs merge=lfs -text +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl filter=lfs diff=lfs merge=lfs -text +*.tar filter=lfs diff=lfs merge=lfs -text +*.wasm filter=lfs diff=lfs merge=lfs -text +*.zst filter=lfs diff=lfs merge=lfs -text +*tfevents* filter=lfs diff=lfs merge=lfs -text + +tokenizer.json filter=lfs diff=lfs merge=lfs -text +training_args.bin filter=lfs diff=lfs merge=lfs -text +vocab.json filter=lfs diff=lfs merge=lfs -text +model.safetensors filter=lfs diff=lfs merge=lfs -text +merges.txt filter=lfs diff=lfs merge=lfs -text \ No newline at end of file diff --git a/README.md b/README.md new file mode 100644 index 0000000..6e7d83e --- /dev/null +++ b/README.md @@ -0,0 +1,56 @@ +--- +library_name: transformers +license: other +base_model: Qwen/Qwen3-1.7B +tags: +- llama-factory +- full +- openthoughts +model-index: +- name: openthoughts-4-code-qwen3-32b-annotated-7k_qwen3-1.7B_10k + results: [] +--- + + +# Openthoughts-4-code-qwen3-32b-annotated-7k_qwen3-1.7B_10k + +This model is a fine-tuned version of [Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B), 10k output length, on the laion/openthoughts-4-code-qwen3-32b-7k-annotated-sharegpt dataset (which is created using https://huggingface.co/datasets/mlfoundations-dev/hero_run_4_code as seed for generation, Qwen3 32B with 7.5k output length) + +## Model description + +More information needed + +## Intended uses & limitations + +More information needed + +## Training and evaluation data + +More information needed + +## Training procedure + +### Training hyperparameters + +The following hyperparameters were used during training: +- learning_rate: 0.0002 +- per_device_train_batch_size: 4 +- seed: 42 +- distributed_type: multi-GPU +- num_devices: 256 +- total_train_batch_size: 1024 +- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.99) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments +- lr_scheduler_type: cosine +- lr_scheduler_warmup_ratio: 0.1 +- num_epochs: 8.0 + +### Training results + + + +### Framework versions + +- Transformers 4.55.0 +- Pytorch 2.7.1+cu128 +- Datasets 3.6.0 +- Tokenizers 0.21.1 diff --git a/added_tokens.json b/added_tokens.json new file mode 100644 index 0000000..b54f913 --- /dev/null +++ b/added_tokens.json @@ -0,0 +1,28 @@ +{ + "": 151668, + "": 151658, + "": 151666, + "": 151667, + "": 151657, + "": 151665, + "<|box_end|>": 151649, + "<|box_start|>": 151648, + "<|endoftext|>": 151643, + "<|file_sep|>": 151664, + "<|fim_middle|>": 151660, + "<|fim_pad|>": 151662, + "<|fim_prefix|>": 151659, + "<|fim_suffix|>": 151661, + "<|im_end|>": 151645, + "<|im_start|>": 151644, + "<|image_pad|>": 151655, + "<|object_ref_end|>": 151647, + "<|object_ref_start|>": 151646, + "<|quad_end|>": 151651, + "<|quad_start|>": 151650, + "<|repo_name|>": 151663, + "<|video_pad|>": 151656, + "<|vision_end|>": 151653, + "<|vision_pad|>": 151654, + "<|vision_start|>": 151652 +} diff --git a/all_results.json b/all_results.json new file mode 100644 index 0000000..bd6d4fd --- /dev/null +++ b/all_results.json @@ -0,0 +1,16 @@ +{ + "achieved_tflops_per_gpu": 0.009154196042913305, + "achieved_tflops_per_gpu_theoretical": 253.1246706838733, + "epoch": 8.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4003600478172302, + "mfu_percent": 0.002934037193241444, + "mfu_percent_theoretical": 81.12970214226708, + "total_flos": 4.678153526260531e+16, + "train_loss": 0.13369933033527026, + "train_runtime": 19962.4709, + "train_samples_per_second": 384.408, + "train_steps_per_second": 0.376, + "valid_targets_mean": 28802.2, + "valid_targets_min": 22042 +} \ No newline at end of file diff --git a/chat_template.jinja b/chat_template.jinja new file mode 100644 index 0000000..01be9b3 --- /dev/null +++ b/chat_template.jinja @@ -0,0 +1,89 @@ +{%- if tools %} + {{- '<|im_start|>system\n' }} + {%- if messages[0].role == 'system' %} + {{- messages[0].content + '\n\n' }} + {%- endif %} + {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }} + {%- for tool in tools %} + {{- "\n" }} + {{- tool | tojson }} + {%- endfor %} + {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }} +{%- else %} + {%- if messages[0].role == 'system' %} + {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }} + {%- endif %} +{%- endif %} +{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %} +{%- for message in messages[::-1] %} + {%- set index = (messages|length - 1) - loop.index0 %} + {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('') and message.content.endswith('')) %} + {%- set ns.multi_step_tool = false %} + {%- set ns.last_query_index = index %} + {%- endif %} +{%- endfor %} +{%- for message in messages %} + {%- if message.content is string %} + {%- set content = message.content %} + {%- else %} + {%- set content = '' %} + {%- endif %} + {%- if (message.role == "user") or (message.role == "system" and not loop.first) %} + {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }} + {%- elif message.role == "assistant" %} + {%- set reasoning_content = '' %} + {%- if message.reasoning_content is string %} + {%- set reasoning_content = message.reasoning_content %} + {%- else %} + {%- if '' in content %} + {%- set reasoning_content = content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %} + {%- set content = content.split('')[-1].lstrip('\n') %} + {%- endif %} + {%- endif %} + {%- if loop.index0 > ns.last_query_index %} + {%- if loop.last or (not loop.last and reasoning_content) %} + {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }} + {%- else %} + {{- '<|im_start|>' + message.role + '\n' + content }} + {%- endif %} + {%- else %} + {{- '<|im_start|>' + message.role + '\n' + content }} + {%- endif %} + {%- if message.tool_calls %} + {%- for tool_call in message.tool_calls %} + {%- if (loop.first and content) or (not loop.first) %} + {{- '\n' }} + {%- endif %} + {%- if tool_call.function %} + {%- set tool_call = tool_call.function %} + {%- endif %} + {{- '\n{"name": "' }} + {{- tool_call.name }} + {{- '", "arguments": ' }} + {%- if tool_call.arguments is string %} + {{- tool_call.arguments }} + {%- else %} + {{- tool_call.arguments | tojson }} + {%- endif %} + {{- '}\n' }} + {%- endfor %} + {%- endif %} + {{- '<|im_end|>\n' }} + {%- elif message.role == "tool" %} + {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %} + {{- '<|im_start|>user' }} + {%- endif %} + {{- '\n\n' }} + {{- content }} + {{- '\n' }} + {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %} + {{- '<|im_end|>\n' }} + {%- endif %} + {%- endif %} +{%- endfor %} +{%- if add_generation_prompt %} + {{- '<|im_start|>assistant\n' }} + {%- if enable_thinking is defined and enable_thinking is false %} + {{- '\n\n\n\n' }} + {%- endif %} +{%- endif %} \ No newline at end of file diff --git a/config.json b/config.json new file mode 100644 index 0000000..26358f4 --- /dev/null +++ b/config.json @@ -0,0 +1,60 @@ +{ + "architectures": [ + "Qwen3ForCausalLM" + ], + "attention_bias": false, + "attention_dropout": 0.0, + "bos_token_id": 151643, + "eos_token_id": 151645, + "head_dim": 128, + "hidden_act": "silu", + "hidden_size": 2048, + "initializer_range": 0.02, + "intermediate_size": 6144, + "layer_types": [ + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention" + ], + "max_position_embeddings": 40960, + "max_window_layers": 28, + "model_type": "qwen3", + "num_attention_heads": 16, + "num_hidden_layers": 28, + "num_key_value_heads": 8, + "rms_norm_eps": 1e-06, + "rope_scaling": null, + "rope_theta": 1000000, + "sliding_window": null, + "tie_word_embeddings": true, + "torch_dtype": "bfloat16", + "transformers_version": "4.55.0", + "use_cache": false, + "use_sliding_window": false, + "vocab_size": 151936 +} diff --git a/configuration.json b/configuration.json new file mode 100644 index 0000000..bbeeda1 --- /dev/null +++ b/configuration.json @@ -0,0 +1 @@ +{"framework": "pytorch", "task": "text-generation", "allow_remote": true} \ No newline at end of file diff --git a/generation_config.json b/generation_config.json new file mode 100644 index 0000000..4b23077 --- /dev/null +++ b/generation_config.json @@ -0,0 +1,13 @@ +{ + "bos_token_id": 151643, + "do_sample": true, + "eos_token_id": [ + 151645, + 151643 + ], + "pad_token_id": 151643, + "temperature": 0.6, + "top_k": 20, + "top_p": 0.95, + "transformers_version": "4.55.0" +} diff --git a/merges.txt b/merges.txt new file mode 100644 index 0000000..80c1a19 --- /dev/null +++ b/merges.txt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8831e4f1a044471340f7c0a83d7bd71306a5b867e95fd870f74d0c5308a904d5 +size 1671853 diff --git a/model.safetensors b/model.safetensors new file mode 100644 index 0000000..fba3259 --- /dev/null +++ b/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b095aafb58d65441a32ad01742c6ebe95b0d28b6f1beb3599548d666978aa66d +size 3441185608 diff --git a/run_summary.json b/run_summary.json new file mode 100644 index 0000000..4308b17 --- /dev/null +++ b/run_summary.json @@ -0,0 +1,12 @@ +{ + "agent_name": null, + "training_start": null, + "training_end": null, + "created_by": "DCAgent", + "base_model_name": "/p/scratch/synthlaion/dc-agent-shared/hf_hub/models--Qwen--Qwen3-1.7B/snapshots/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e", + "dataset_name": "laion/openthoughts-4-code-qwen3-32b-7k-annotated-sharegpt", + "training_type": "SFT", + "training_parameters": "https://huggingface.co/mlfoundations-dev/openthoughts-4-code-qwen3-32b-7k-annotated-sharegpt_save-strategy_steps_Qwen3-1-7B/blob/main/config.json", + "wandb_link": null, + "traces_location_s3": null +} \ No newline at end of file diff --git a/special_tokens_map.json b/special_tokens_map.json new file mode 100644 index 0000000..ac23c0a --- /dev/null +++ b/special_tokens_map.json @@ -0,0 +1,31 @@ +{ + "additional_special_tokens": [ + "<|im_start|>", + "<|im_end|>", + "<|object_ref_start|>", + "<|object_ref_end|>", + "<|box_start|>", + "<|box_end|>", + "<|quad_start|>", + "<|quad_end|>", + "<|vision_start|>", + "<|vision_end|>", + "<|vision_pad|>", + "<|image_pad|>", + "<|video_pad|>" + ], + "eos_token": { + "content": "<|im_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false + }, + "pad_token": { + "content": "<|endoftext|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false + } +} diff --git a/tokenizer.json b/tokenizer.json new file mode 100644 index 0000000..cd71f61 --- /dev/null +++ b/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4 +size 11422654 diff --git a/tokenizer_config.json b/tokenizer_config.json new file mode 100644 index 0000000..a6ef16f --- /dev/null +++ b/tokenizer_config.json @@ -0,0 +1,240 @@ +{ + "add_bos_token": false, + "add_prefix_space": false, + "added_tokens_decoder": { + "151643": { + "content": "<|endoftext|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151644": { + "content": "<|im_start|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151645": { + "content": "<|im_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151646": { + "content": "<|object_ref_start|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151647": { + "content": "<|object_ref_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151648": { + "content": "<|box_start|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151649": { + "content": "<|box_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151650": { + "content": "<|quad_start|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151651": { + "content": "<|quad_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151652": { + "content": "<|vision_start|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151653": { + "content": "<|vision_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151654": { + "content": "<|vision_pad|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151655": { + "content": "<|image_pad|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151656": { + "content": "<|video_pad|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "151657": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151658": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151659": { + "content": "<|fim_prefix|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151660": { + "content": "<|fim_middle|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151661": { + "content": "<|fim_suffix|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151662": { + "content": "<|fim_pad|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151663": { + "content": "<|repo_name|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151664": { + "content": "<|file_sep|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151665": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151666": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151667": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "151668": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + } + }, + "additional_special_tokens": [ + "<|im_start|>", + "<|im_end|>", + "<|object_ref_start|>", + "<|object_ref_end|>", + "<|box_start|>", + "<|box_end|>", + "<|quad_start|>", + "<|quad_end|>", + "<|vision_start|>", + "<|vision_end|>", + "<|vision_pad|>", + "<|image_pad|>", + "<|video_pad|>" + ], + "bos_token": null, + "clean_up_tokenization_spaces": false, + "eos_token": "<|im_end|>", + "errors": "replace", + "extra_special_tokens": {}, + "model_max_length": 10240, + "pad_token": "<|endoftext|>", + "padding_side": "right", + "split_special_tokens": false, + "tokenizer_class": "Qwen2Tokenizer", + "unk_token": null +} diff --git a/train_results.json b/train_results.json new file mode 100644 index 0000000..bd6d4fd --- /dev/null +++ b/train_results.json @@ -0,0 +1,16 @@ +{ + "achieved_tflops_per_gpu": 0.009154196042913305, + "achieved_tflops_per_gpu_theoretical": 253.1246706838733, + "epoch": 8.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4003600478172302, + "mfu_percent": 0.002934037193241444, + "mfu_percent_theoretical": 81.12970214226708, + "total_flos": 4.678153526260531e+16, + "train_loss": 0.13369933033527026, + "train_runtime": 19962.4709, + "train_samples_per_second": 384.408, + "train_steps_per_second": 0.376, + "valid_targets_mean": 28802.2, + "valid_targets_min": 22042 +} \ No newline at end of file diff --git a/trainer_log.jsonl b/trainer_log.jsonl new file mode 100644 index 0000000..c3924ca --- /dev/null +++ b/trainer_log.jsonl @@ -0,0 +1,1614 @@ +{"current_steps": 5, "total_steps": 59952, "loss": 0.8079, "lr": 1.3342228152101402e-07, "epoch": 0.00016680288902603792, "percentage": 0.01, "elapsed_time": "0:00:52", "remaining_time": "7 days, 7:31:15"} +{"current_steps": 10, "total_steps": 59952, "loss": 0.8181, "lr": 3.002001334222815e-07, "epoch": 0.00033360577805207584, "percentage": 0.02, "elapsed_time": "0:01:25", "remaining_time": "5 days, 22:58:44"} +{"current_steps": 15, "total_steps": 59952, "loss": 0.8136, "lr": 4.669779853235491e-07, "epoch": 0.0005004086670781138, "percentage": 0.03, "elapsed_time": "0:01:56", "remaining_time": "5 days, 9:21:23"} +{"current_steps": 20, "total_steps": 59952, "loss": 0.7905, "lr": 6.337558372248166e-07, "epoch": 0.0006672115561041517, "percentage": 0.03, "elapsed_time": "0:02:26", "remaining_time": "5 days, 2:14:24"} +{"current_steps": 25, "total_steps": 59952, "loss": 0.7999, "lr": 8.005336891260842e-07, "epoch": 0.0008340144451301897, "percentage": 0.04, "elapsed_time": "0:02:55", "remaining_time": "4 days, 21:05:29"} +{"current_steps": 30, "total_steps": 59952, "loss": 0.7699, "lr": 9.673115410273516e-07, "epoch": 0.0010008173341562276, "percentage": 0.05, "elapsed_time": "0:03:26", "remaining_time": "4 days, 18:17:52"} +{"current_steps": 35, "total_steps": 59952, "loss": 0.7695, "lr": 1.134089392928619e-06, "epoch": 0.0011676202231822655, "percentage": 0.06, "elapsed_time": "0:03:59", "remaining_time": "4 days, 17:44:40"} +{"current_steps": 40, "total_steps": 59952, "loss": 0.747, "lr": 1.3008672448298865e-06, "epoch": 0.0013344231122083034, "percentage": 0.07, "elapsed_time": "0:04:29", "remaining_time": "4 days, 16:05:18"} +{"current_steps": 45, "total_steps": 59952, "loss": 0.7426, "lr": 1.4676450967311542e-06, "epoch": 0.0015012260012343414, "percentage": 0.08, "elapsed_time": "0:05:00", "remaining_time": "4 days, 15:12:34"} +{"current_steps": 50, "total_steps": 59952, "loss": 0.6996, "lr": 1.6344229486324217e-06, "epoch": 0.0016680288902603793, "percentage": 0.08, "elapsed_time": "0:05:30", "remaining_time": "4 days, 14:00:46"} +{"current_steps": 55, "total_steps": 59952, "loss": 0.6994, "lr": 1.8012008005336891e-06, "epoch": 0.0018348317792864172, "percentage": 0.09, "elapsed_time": "0:06:01", "remaining_time": "4 days, 13:16:05"} +{"current_steps": 60, "total_steps": 59952, "loss": 0.6791, "lr": 1.9679786524349566e-06, "epoch": 0.0020016346683124553, "percentage": 0.1, "elapsed_time": "0:06:30", "remaining_time": "4 days, 12:22:13"} +{"current_steps": 65, "total_steps": 59952, "loss": 0.6699, "lr": 2.1347565043362243e-06, "epoch": 0.002168437557338493, "percentage": 0.11, "elapsed_time": "0:07:00", "remaining_time": "4 days, 11:31:05"} +{"current_steps": 70, "total_steps": 59952, "loss": 0.6618, "lr": 2.301534356237492e-06, "epoch": 0.002335240446364531, "percentage": 0.12, "elapsed_time": "0:07:30", "remaining_time": "4 days, 11:00:11"} +{"current_steps": 75, "total_steps": 59952, "loss": 0.6562, "lr": 2.468312208138759e-06, "epoch": 0.002502043335390569, "percentage": 0.13, "elapsed_time": "0:08:00", "remaining_time": "4 days, 10:38:17"} +{"current_steps": 80, "total_steps": 59952, "loss": 0.651, "lr": 2.635090060040027e-06, "epoch": 0.0026688462244166067, "percentage": 0.13, "elapsed_time": "0:08:30", "remaining_time": "4 days, 10:04:47"} +{"current_steps": 5, "total_steps": 59952, "loss": 0.8079, "lr": 1.3342228152101402e-07, "epoch": 0.00016680288902603792, "percentage": 0.01, "elapsed_time": "0:00:51", "remaining_time": "7 days, 3:37:50"} +{"current_steps": 5, "total_steps": 59952, "loss": 0.8079, "lr": 1.3342228152101402e-07, "epoch": 0.00016680288902603792, "percentage": 0.01, "elapsed_time": "0:00:51", "remaining_time": "7 days, 1:56:31"} +{"current_steps": 10, "total_steps": 59952, "loss": 0.8181, "lr": 3.002001334222815e-07, "epoch": 0.00033360577805207584, "percentage": 0.02, "elapsed_time": "0:01:24", "remaining_time": "5 days, 20:30:41"} +{"current_steps": 15, "total_steps": 59952, "loss": 0.8136, "lr": 4.669779853235491e-07, "epoch": 0.0005004086670781138, "percentage": 0.03, "elapsed_time": "0:01:55", "remaining_time": "5 days, 8:18:57"} +{"current_steps": 20, "total_steps": 59952, "loss": 0.7904, "lr": 6.337558372248166e-07, "epoch": 0.0006672115561041517, "percentage": 0.03, "elapsed_time": "0:02:25", "remaining_time": "5 days, 1:29:25"} +{"current_steps": 25, "total_steps": 59952, "loss": 0.7999, "lr": 8.005336891260842e-07, "epoch": 0.0008340144451301897, "percentage": 0.04, "elapsed_time": "0:02:54", "remaining_time": "4 days, 20:29:41"} +{"current_steps": 30, "total_steps": 59952, "loss": 0.77, "lr": 9.673115410273516e-07, "epoch": 0.0010008173341562276, "percentage": 0.05, "elapsed_time": "0:03:26", "remaining_time": "4 days, 18:19:14"} +{"current_steps": 35, "total_steps": 59952, "loss": 0.7695, "lr": 1.134089392928619e-06, "epoch": 0.0011676202231822655, "percentage": 0.06, "elapsed_time": "0:03:59", "remaining_time": "4 days, 17:59:58"} +{"current_steps": 40, "total_steps": 59952, "loss": 0.747, "lr": 1.3008672448298865e-06, "epoch": 0.0013344231122083034, "percentage": 0.07, "elapsed_time": "0:04:30", "remaining_time": "4 days, 16:44:01"} +{"current_steps": 45, "total_steps": 59952, "loss": 0.7426, "lr": 1.4676450967311542e-06, "epoch": 0.0015012260012343414, "percentage": 0.08, "elapsed_time": "0:05:02", "remaining_time": "4 days, 15:40:46"} +{"current_steps": 50, "total_steps": 59952, "loss": 0.6996, "lr": 1.6344229486324217e-06, "epoch": 0.0016680288902603793, "percentage": 0.08, "elapsed_time": "0:05:31", "remaining_time": "4 days, 14:23:47"} +{"current_steps": 55, "total_steps": 59952, "loss": 0.6995, "lr": 1.8012008005336891e-06, "epoch": 0.0018348317792864172, "percentage": 0.09, "elapsed_time": "0:06:02", "remaining_time": "4 days, 13:35:27"} +{"current_steps": 60, "total_steps": 59952, "loss": 0.6791, "lr": 1.9679786524349566e-06, "epoch": 0.0020016346683124553, "percentage": 0.1, "elapsed_time": "0:06:31", "remaining_time": "4 days, 12:41:28"} +{"current_steps": 65, "total_steps": 59952, "loss": 0.67, "lr": 2.1347565043362243e-06, "epoch": 0.002168437557338493, "percentage": 0.11, "elapsed_time": "0:07:01", "remaining_time": "4 days, 11:46:38"} +{"current_steps": 70, "total_steps": 59952, "loss": 0.6618, "lr": 2.301534356237492e-06, "epoch": 0.002335240446364531, "percentage": 0.12, "elapsed_time": "0:07:31", "remaining_time": "4 days, 11:10:13"} +{"current_steps": 75, "total_steps": 59952, "loss": 0.6563, "lr": 2.468312208138759e-06, "epoch": 0.002502043335390569, "percentage": 0.13, "elapsed_time": "0:08:02", "remaining_time": "4 days, 10:53:50"} +{"current_steps": 5, "total_steps": 59952, "loss": 0.8079, "lr": 1.3342228152101402e-07, "epoch": 0.00016680010675206833, "percentage": 0.01, "elapsed_time": "0:00:37", "remaining_time": "5 days, 6:13:02"} +{"current_steps": 10, "total_steps": 59952, "loss": 0.8182, "lr": 3.002001334222815e-07, "epoch": 0.00033360021350413666, "percentage": 0.02, "elapsed_time": "0:00:57", "remaining_time": "4 days, 0:29:29"} +{"current_steps": 15, "total_steps": 59952, "loss": 0.8135, "lr": 4.669779853235491e-07, "epoch": 0.0005004003202562049, "percentage": 0.03, "elapsed_time": "0:01:17", "remaining_time": "3 days, 13:39:53"} +{"current_steps": 20, "total_steps": 59952, "loss": 0.7904, "lr": 6.337558372248166e-07, "epoch": 0.0006672004270082733, "percentage": 0.03, "elapsed_time": "0:01:35", "remaining_time": "3 days, 7:42:56"} +{"current_steps": 25, "total_steps": 59952, "loss": 0.7999, "lr": 8.005336891260842e-07, "epoch": 0.0008340005337603416, "percentage": 0.04, "elapsed_time": "0:01:53", "remaining_time": "3 days, 3:43:40"} +{"current_steps": 30, "total_steps": 59952, "loss": 0.77, "lr": 9.673115410273516e-07, "epoch": 0.0010008006405124099, "percentage": 0.05, "elapsed_time": "0:02:13", "remaining_time": "3 days, 1:53:21"} +{"current_steps": 35, "total_steps": 59952, "loss": 0.7695, "lr": 1.134089392928619e-06, "epoch": 0.0011676007472644784, "percentage": 0.06, "elapsed_time": "0:02:31", "remaining_time": "2 days, 23:57:10"} +{"current_steps": 40, "total_steps": 59952, "loss": 0.747, "lr": 1.3008672448298865e-06, "epoch": 0.0013344008540165466, "percentage": 0.07, "elapsed_time": "0:02:49", "remaining_time": "2 days, 22:36:34"} +{"current_steps": 45, "total_steps": 59952, "loss": 0.7426, "lr": 1.4676450967311542e-06, "epoch": 0.001501200960768615, "percentage": 0.08, "elapsed_time": "0:03:09", "remaining_time": "2 days, 21:59:04"} +{"current_steps": 50, "total_steps": 59952, "loss": 0.6995, "lr": 1.6344229486324217e-06, "epoch": 0.0016680010675206832, "percentage": 0.08, "elapsed_time": "0:03:27", "remaining_time": "2 days, 21:07:10"} +{"current_steps": 55, "total_steps": 59952, "loss": 0.6995, "lr": 1.8012008005336891e-06, "epoch": 0.0018348011742727514, "percentage": 0.09, "elapsed_time": "0:03:45", "remaining_time": "2 days, 20:20:18"} +{"current_steps": 60, "total_steps": 59952, "loss": 0.6791, "lr": 1.9679786524349566e-06, "epoch": 0.0020016012810248197, "percentage": 0.1, "elapsed_time": "0:04:04", "remaining_time": "2 days, 19:52:03"} +{"current_steps": 65, "total_steps": 59952, "loss": 0.6699, "lr": 2.1347565043362243e-06, "epoch": 0.002168401387776888, "percentage": 0.11, "elapsed_time": "0:04:23", "remaining_time": "2 days, 19:19:32"} +{"current_steps": 70, "total_steps": 59952, "loss": 0.6618, "lr": 2.301534356237492e-06, "epoch": 0.0023352014945289567, "percentage": 0.12, "elapsed_time": "0:04:41", "remaining_time": "2 days, 18:51:17"} +{"current_steps": 75, "total_steps": 59952, "loss": 0.6562, "lr": 2.468312208138759e-06, "epoch": 0.0025020016012810248, "percentage": 0.13, "elapsed_time": "0:05:00", "remaining_time": "2 days, 18:38:02"} +{"current_steps": 80, "total_steps": 59952, "loss": 0.6509, "lr": 2.635090060040027e-06, "epoch": 0.0026688017080330933, "percentage": 0.13, "elapsed_time": "0:05:18", "remaining_time": "2 days, 18:17:41"} +{"current_steps": 85, "total_steps": 59952, "loss": 0.6507, "lr": 2.8018679119412946e-06, "epoch": 0.0028356018147851613, "percentage": 0.14, "elapsed_time": "0:05:37", "remaining_time": "2 days, 18:05:04"} +{"current_steps": 90, "total_steps": 59952, "loss": 0.6512, "lr": 2.968645763842562e-06, "epoch": 0.00300240192153723, "percentage": 0.15, "elapsed_time": "0:05:55", "remaining_time": "2 days, 17:42:18"} +{"current_steps": 95, "total_steps": 59952, "loss": 0.6118, "lr": 3.1354236157438295e-06, "epoch": 0.0031692020282892983, "percentage": 0.16, "elapsed_time": "0:06:13", "remaining_time": "2 days, 17:24:13"} +{"current_steps": 100, "total_steps": 59952, "loss": 0.639, "lr": 3.3022014676450967e-06, "epoch": 0.0033360021350413663, "percentage": 0.17, "elapsed_time": "0:06:31", "remaining_time": "2 days, 17:07:15"} +{"current_steps": 105, "total_steps": 59952, "loss": 0.6146, "lr": 3.4689793195463644e-06, "epoch": 0.003502802241793435, "percentage": 0.18, "elapsed_time": "0:06:57", "remaining_time": "2 days, 18:06:10"} +{"current_steps": 110, "total_steps": 59952, "loss": 0.6219, "lr": 3.6357571714476317e-06, "epoch": 0.003669602348545503, "percentage": 0.18, "elapsed_time": "0:07:15", "remaining_time": "2 days, 17:47:14"} +{"current_steps": 115, "total_steps": 59952, "loss": 0.6218, "lr": 3.8025350233488993e-06, "epoch": 0.0038364024552975714, "percentage": 0.19, "elapsed_time": "0:07:33", "remaining_time": "2 days, 17:30:08"} +{"current_steps": 120, "total_steps": 59952, "loss": 0.6231, "lr": 3.969312875250167e-06, "epoch": 0.0040032025620496394, "percentage": 0.2, "elapsed_time": "0:07:51", "remaining_time": "2 days, 17:14:40"} +{"current_steps": 125, "total_steps": 59952, "loss": 0.6134, "lr": 4.136090727151434e-06, "epoch": 0.004170002668801708, "percentage": 0.21, "elapsed_time": "0:08:08", "remaining_time": "2 days, 17:00:16"} +{"current_steps": 130, "total_steps": 59952, "loss": 0.6159, "lr": 4.302868579052702e-06, "epoch": 0.004336802775553776, "percentage": 0.22, "elapsed_time": "0:08:26", "remaining_time": "2 days, 16:46:44"} +{"current_steps": 135, "total_steps": 59952, "loss": 0.6321, "lr": 4.46964643095397e-06, "epoch": 0.0045036028823058445, "percentage": 0.23, "elapsed_time": "0:08:44", "remaining_time": "2 days, 16:34:19"} +{"current_steps": 140, "total_steps": 59952, "loss": 0.618, "lr": 4.6364242828552364e-06, "epoch": 0.004670402989057913, "percentage": 0.23, "elapsed_time": "0:09:02", "remaining_time": "2 days, 16:24:04"} +{"current_steps": 145, "total_steps": 59952, "loss": 0.6029, "lr": 4.803202134756504e-06, "epoch": 0.0048372030958099815, "percentage": 0.24, "elapsed_time": "0:09:20", "remaining_time": "2 days, 16:12:54"} +{"current_steps": 150, "total_steps": 59952, "loss": 0.6325, "lr": 4.969979986657772e-06, "epoch": 0.0050040032025620495, "percentage": 0.25, "elapsed_time": "0:09:38", "remaining_time": "2 days, 16:03:56"} +{"current_steps": 155, "total_steps": 59952, "loss": 0.6183, "lr": 5.13675783855904e-06, "epoch": 0.005170803309314118, "percentage": 0.26, "elapsed_time": "0:09:56", "remaining_time": "2 days, 15:54:09"} +{"current_steps": 160, "total_steps": 59952, "loss": 0.6183, "lr": 5.303535690460307e-06, "epoch": 0.0053376034160661865, "percentage": 0.27, "elapsed_time": "0:10:14", "remaining_time": "2 days, 15:47:39"} +{"current_steps": 165, "total_steps": 59952, "loss": 0.615, "lr": 5.470313542361575e-06, "epoch": 0.0055044035228182546, "percentage": 0.28, "elapsed_time": "0:10:32", "remaining_time": "2 days, 15:41:33"} +{"current_steps": 170, "total_steps": 59952, "loss": 0.6037, "lr": 5.6370913942628425e-06, "epoch": 0.005671203629570323, "percentage": 0.28, "elapsed_time": "0:10:50", "remaining_time": "2 days, 15:33:49"} +{"current_steps": 175, "total_steps": 59952, "loss": 0.6152, "lr": 5.80386924616411e-06, "epoch": 0.0058380037363223915, "percentage": 0.29, "elapsed_time": "0:11:08", "remaining_time": "2 days, 15:26:12"} +{"current_steps": 180, "total_steps": 59952, "loss": 0.6304, "lr": 5.970647098065377e-06, "epoch": 0.00600480384307446, "percentage": 0.3, "elapsed_time": "0:11:26", "remaining_time": "2 days, 15:18:55"} +{"current_steps": 185, "total_steps": 59952, "loss": 0.6191, "lr": 6.137424949966645e-06, "epoch": 0.006171603949826528, "percentage": 0.31, "elapsed_time": "0:11:44", "remaining_time": "2 days, 15:12:04"} +{"current_steps": 190, "total_steps": 59952, "loss": 0.6102, "lr": 6.3042028018679115e-06, "epoch": 0.006338404056578597, "percentage": 0.32, "elapsed_time": "0:12:02", "remaining_time": "2 days, 15:05:36"} +{"current_steps": 195, "total_steps": 59952, "loss": 0.597, "lr": 6.470980653769179e-06, "epoch": 0.006505204163330665, "percentage": 0.33, "elapsed_time": "0:12:19", "remaining_time": "2 days, 14:59:25"} +{"current_steps": 200, "total_steps": 59952, "loss": 0.6139, "lr": 6.637758505670447e-06, "epoch": 0.006672004270082733, "percentage": 0.33, "elapsed_time": "0:12:38", "remaining_time": "2 days, 14:56:19"} +{"current_steps": 205, "total_steps": 59952, "loss": 0.6073, "lr": 6.8045363575717145e-06, "epoch": 0.006838804376834801, "percentage": 0.34, "elapsed_time": "0:13:03", "remaining_time": "2 days, 15:23:43"} +{"current_steps": 210, "total_steps": 59952, "loss": 0.5927, "lr": 6.971314209472981e-06, "epoch": 0.00700560448358687, "percentage": 0.35, "elapsed_time": "0:13:20", "remaining_time": "2 days, 15:17:24"} +{"current_steps": 215, "total_steps": 59952, "loss": 0.5971, "lr": 7.138092061374251e-06, "epoch": 0.007172404590338938, "percentage": 0.36, "elapsed_time": "0:13:38", "remaining_time": "2 days, 15:11:31"} +{"current_steps": 220, "total_steps": 59952, "loss": 0.5964, "lr": 7.304869913275518e-06, "epoch": 0.007339204697091006, "percentage": 0.37, "elapsed_time": "0:13:56", "remaining_time": "2 days, 15:05:49"} +{"current_steps": 225, "total_steps": 59952, "loss": 0.6138, "lr": 7.471647765176785e-06, "epoch": 0.007506004803843075, "percentage": 0.38, "elapsed_time": "0:14:14", "remaining_time": "2 days, 15:00:08"} +{"current_steps": 230, "total_steps": 59952, "loss": 0.5926, "lr": 7.638425617078054e-06, "epoch": 0.007672804910595143, "percentage": 0.38, "elapsed_time": "0:14:32", "remaining_time": "2 days, 14:54:53"} +{"current_steps": 235, "total_steps": 59952, "loss": 0.6271, "lr": 7.80520346897932e-06, "epoch": 0.007839605017347212, "percentage": 0.39, "elapsed_time": "0:14:50", "remaining_time": "2 days, 14:52:24"} +{"current_steps": 240, "total_steps": 59952, "loss": 0.6018, "lr": 7.971981320880587e-06, "epoch": 0.008006405124099279, "percentage": 0.4, "elapsed_time": "0:15:09", "remaining_time": "2 days, 14:49:28"} +{"current_steps": 245, "total_steps": 59952, "loss": 0.6036, "lr": 8.138759172781855e-06, "epoch": 0.008173205230851348, "percentage": 0.41, "elapsed_time": "0:15:27", "remaining_time": "2 days, 14:46:05"} +{"current_steps": 250, "total_steps": 59952, "loss": 0.6236, "lr": 8.305537024683123e-06, "epoch": 0.008340005337603417, "percentage": 0.42, "elapsed_time": "0:15:45", "remaining_time": "2 days, 14:41:17"} +{"current_steps": 255, "total_steps": 59952, "loss": 0.6075, "lr": 8.47231487658439e-06, "epoch": 0.008506805444355484, "percentage": 0.43, "elapsed_time": "0:16:02", "remaining_time": "2 days, 14:36:50"} +{"current_steps": 260, "total_steps": 59952, "loss": 0.6203, "lr": 8.639092728485658e-06, "epoch": 0.008673605551107553, "percentage": 0.43, "elapsed_time": "0:16:21", "remaining_time": "2 days, 14:34:40"} +{"current_steps": 265, "total_steps": 59952, "loss": 0.6098, "lr": 8.805870580386926e-06, "epoch": 0.008840405657859622, "percentage": 0.44, "elapsed_time": "0:16:39", "remaining_time": "2 days, 14:30:22"} +{"current_steps": 270, "total_steps": 59952, "loss": 0.5909, "lr": 8.972648432288193e-06, "epoch": 0.009007205764611689, "percentage": 0.45, "elapsed_time": "0:16:56", "remaining_time": "2 days, 14:26:20"} +{"current_steps": 275, "total_steps": 59952, "loss": 0.5765, "lr": 9.139426284189461e-06, "epoch": 0.009174005871363758, "percentage": 0.46, "elapsed_time": "0:17:15", "remaining_time": "2 days, 14:24:44"} +{"current_steps": 280, "total_steps": 59952, "loss": 0.5921, "lr": 9.306204136090727e-06, "epoch": 0.009340805978115827, "percentage": 0.47, "elapsed_time": "0:17:33", "remaining_time": "2 days, 14:21:09"} +{"current_steps": 285, "total_steps": 59952, "loss": 0.5947, "lr": 9.472981987991995e-06, "epoch": 0.009507606084867894, "percentage": 0.48, "elapsed_time": "0:17:51", "remaining_time": "2 days, 14:17:29"} +{"current_steps": 5, "total_steps": 7496, "loss": 0.8061, "lr": 1.0666666666666667e-06, "epoch": 0.005336179295624333, "percentage": 0.07, "elapsed_time": "0:11:33", "remaining_time": "12 days, 0:27:04"} +{"current_steps": 5, "total_steps": 7496, "loss": 0.8061, "lr": 1.0666666666666667e-06, "epoch": 0.005336179295624333, "percentage": 0.07, "elapsed_time": "0:16:57", "remaining_time": "17 days, 15:37:31"} +{"current_steps": 10, "total_steps": 7496, "loss": 0.7944, "lr": 2.4000000000000003e-06, "epoch": 0.010672358591248666, "percentage": 0.13, "elapsed_time": "0:17:21", "remaining_time": "9 days, 0:34:36"} +{"current_steps": 15, "total_steps": 7496, "loss": 0.7523, "lr": 3.7333333333333337e-06, "epoch": 0.016008537886873, "percentage": 0.2, "elapsed_time": "0:17:49", "remaining_time": "6 days, 4:09:24"} +{"current_steps": 20, "total_steps": 7496, "loss": 0.7026, "lr": 5.066666666666667e-06, "epoch": 0.021344717182497332, "percentage": 0.27, "elapsed_time": "0:18:45", "remaining_time": "4 days, 20:52:11"} +{"current_steps": 25, "total_steps": 7496, "loss": 0.6692, "lr": 6.4000000000000006e-06, "epoch": 0.026680896478121666, "percentage": 0.33, "elapsed_time": "0:19:09", "remaining_time": "3 days, 23:26:53"} +{"current_steps": 30, "total_steps": 7496, "loss": 0.6579, "lr": 7.733333333333334e-06, "epoch": 0.032017075773746, "percentage": 0.4, "elapsed_time": "0:19:35", "remaining_time": "3 days, 9:15:15"} +{"current_steps": 35, "total_steps": 7496, "loss": 0.6443, "lr": 9.066666666666667e-06, "epoch": 0.03735325506937033, "percentage": 0.47, "elapsed_time": "0:20:31", "remaining_time": "3 days, 0:54:27"} +{"current_steps": 40, "total_steps": 7496, "loss": 0.6329, "lr": 1.04e-05, "epoch": 0.042689434364994665, "percentage": 0.53, "elapsed_time": "0:20:53", "remaining_time": "2 days, 16:54:27"} +{"current_steps": 45, "total_steps": 7496, "loss": 0.6266, "lr": 1.1733333333333333e-05, "epoch": 0.048025613660619, "percentage": 0.6, "elapsed_time": "0:21:16", "remaining_time": "2 days, 10:42:53"} +{"current_steps": 50, "total_steps": 7496, "loss": 0.6208, "lr": 1.3066666666666666e-05, "epoch": 0.05336179295624333, "percentage": 0.67, "elapsed_time": "0:22:05", "remaining_time": "2 days, 6:48:42"} +{"current_steps": 55, "total_steps": 7496, "loss": 0.6136, "lr": 1.44e-05, "epoch": 0.05869797225186766, "percentage": 0.73, "elapsed_time": "0:22:34", "remaining_time": "2 days, 2:54:14"} +{"current_steps": 60, "total_steps": 7496, "loss": 0.6083, "lr": 1.5733333333333334e-05, "epoch": 0.064034151547492, "percentage": 0.8, "elapsed_time": "0:22:57", "remaining_time": "1 day, 23:24:43"} +{"current_steps": 65, "total_steps": 7496, "loss": 0.607, "lr": 1.7066666666666667e-05, "epoch": 0.06937033084311633, "percentage": 0.87, "elapsed_time": "0:23:37", "remaining_time": "1 day, 21:01:36"} +{"current_steps": 70, "total_steps": 7496, "loss": 0.6034, "lr": 1.84e-05, "epoch": 0.07470651013874066, "percentage": 0.93, "elapsed_time": "0:24:24", "remaining_time": "1 day, 19:08:45"} +{"current_steps": 75, "total_steps": 7496, "loss": 0.5973, "lr": 1.9733333333333333e-05, "epoch": 0.08004268943436499, "percentage": 1.0, "elapsed_time": "0:24:46", "remaining_time": "1 day, 16:51:53"} +{"current_steps": 80, "total_steps": 7496, "loss": 0.5912, "lr": 2.106666666666667e-05, "epoch": 0.08537886872998933, "percentage": 1.07, "elapsed_time": "0:25:16", "remaining_time": "1 day, 15:03:44"} +{"current_steps": 85, "total_steps": 7496, "loss": 0.5886, "lr": 2.2400000000000002e-05, "epoch": 0.09071504802561366, "percentage": 1.13, "elapsed_time": "0:26:11", "remaining_time": "1 day, 14:04:07"} +{"current_steps": 90, "total_steps": 7496, "loss": 0.5903, "lr": 2.3733333333333335e-05, "epoch": 0.096051227321238, "percentage": 1.2, "elapsed_time": "0:26:35", "remaining_time": "1 day, 12:28:37"} +{"current_steps": 95, "total_steps": 7496, "loss": 0.5842, "lr": 2.5066666666666665e-05, "epoch": 0.10138740661686232, "percentage": 1.27, "elapsed_time": "0:26:58", "remaining_time": "1 day, 11:00:53"} +{"current_steps": 100, "total_steps": 7496, "loss": 0.5839, "lr": 2.64e-05, "epoch": 0.10672358591248667, "percentage": 1.33, "elapsed_time": "0:27:49", "remaining_time": "1 day, 10:17:27"} +{"current_steps": 105, "total_steps": 7496, "loss": 0.5835, "lr": 2.7733333333333334e-05, "epoch": 0.11205976520811099, "percentage": 1.4, "elapsed_time": "0:28:31", "remaining_time": "1 day, 9:27:47"} +{"current_steps": 110, "total_steps": 7496, "loss": 0.5798, "lr": 2.906666666666667e-05, "epoch": 0.11739594450373532, "percentage": 1.47, "elapsed_time": "0:28:54", "remaining_time": "1 day, 8:20:47"} +{"current_steps": 115, "total_steps": 7496, "loss": 0.5784, "lr": 3.04e-05, "epoch": 0.12273212379935966, "percentage": 1.53, "elapsed_time": "0:29:38", "remaining_time": "1 day, 7:42:15"} +{"current_steps": 120, "total_steps": 7496, "loss": 0.5753, "lr": 3.173333333333334e-05, "epoch": 0.128068303094984, "percentage": 1.6, "elapsed_time": "0:30:20", "remaining_time": "1 day, 7:04:39"} +{"current_steps": 125, "total_steps": 7496, "loss": 0.5753, "lr": 3.3066666666666666e-05, "epoch": 0.13340448239060831, "percentage": 1.67, "elapsed_time": "0:30:44", "remaining_time": "1 day, 6:12:19"} +{"current_steps": 130, "total_steps": 7496, "loss": 0.5761, "lr": 3.4399999999999996e-05, "epoch": 0.13874066168623267, "percentage": 1.73, "elapsed_time": "0:31:18", "remaining_time": "1 day, 5:33:37"} +{"current_steps": 135, "total_steps": 7496, "loss": 0.576, "lr": 3.573333333333333e-05, "epoch": 0.144076840981857, "percentage": 1.8, "elapsed_time": "0:32:06", "remaining_time": "1 day, 5:10:50"} +{"current_steps": 140, "total_steps": 7496, "loss": 0.5715, "lr": 3.706666666666667e-05, "epoch": 0.14941302027748132, "percentage": 1.87, "elapsed_time": "0:32:32", "remaining_time": "1 day, 4:29:57"} +{"current_steps": 145, "total_steps": 7496, "loss": 0.5701, "lr": 3.8400000000000005e-05, "epoch": 0.15474919957310565, "percentage": 1.93, "elapsed_time": "0:33:00", "remaining_time": "1 day, 3:53:31"} +{"current_steps": 150, "total_steps": 7496, "loss": 0.5685, "lr": 3.9733333333333335e-05, "epoch": 0.16008537886872998, "percentage": 2.0, "elapsed_time": "0:33:50", "remaining_time": "1 day, 3:36:56"} +{"current_steps": 155, "total_steps": 7496, "loss": 0.5692, "lr": 4.106666666666667e-05, "epoch": 0.16542155816435433, "percentage": 2.07, "elapsed_time": "0:34:22", "remaining_time": "1 day, 3:07:48"} +{"current_steps": 160, "total_steps": 7496, "loss": 0.5663, "lr": 4.24e-05, "epoch": 0.17075773745997866, "percentage": 2.13, "elapsed_time": "0:34:45", "remaining_time": "1 day, 2:33:48"} +{"current_steps": 165, "total_steps": 7496, "loss": 0.5681, "lr": 4.373333333333334e-05, "epoch": 0.17609391675560299, "percentage": 2.2, "elapsed_time": "0:35:30", "remaining_time": "1 day, 2:17:52"} +{"current_steps": 170, "total_steps": 7496, "loss": 0.5683, "lr": 4.5066666666666667e-05, "epoch": 0.1814300960512273, "percentage": 2.27, "elapsed_time": "0:36:09", "remaining_time": "1 day, 1:58:16"} +{"current_steps": 175, "total_steps": 7496, "loss": 0.5667, "lr": 4.64e-05, "epoch": 0.18676627534685167, "percentage": 2.33, "elapsed_time": "0:36:36", "remaining_time": "1 day, 1:31:28"} +{"current_steps": 180, "total_steps": 7496, "loss": 0.5643, "lr": 4.773333333333333e-05, "epoch": 0.192102454642476, "percentage": 2.4, "elapsed_time": "0:37:12", "remaining_time": "1 day, 1:12:23"} +{"current_steps": 185, "total_steps": 7496, "loss": 0.5669, "lr": 4.906666666666667e-05, "epoch": 0.19743863393810032, "percentage": 2.47, "elapsed_time": "0:37:54", "remaining_time": "1 day, 0:57:54"} +{"current_steps": 190, "total_steps": 7496, "loss": 0.5667, "lr": 5.0400000000000005e-05, "epoch": 0.20277481323372465, "percentage": 2.53, "elapsed_time": "0:38:24", "remaining_time": "1 day, 0:36:53"} +{"current_steps": 195, "total_steps": 7496, "loss": 0.5628, "lr": 5.1733333333333335e-05, "epoch": 0.20811099252934898, "percentage": 2.6, "elapsed_time": "0:38:56", "remaining_time": "1 day, 0:18:11"} +{"current_steps": 200, "total_steps": 7496, "loss": 0.5646, "lr": 5.3066666666666665e-05, "epoch": 0.21344717182497333, "percentage": 2.67, "elapsed_time": "0:39:41", "remaining_time": "1 day, 0:07:42"} +{"current_steps": 205, "total_steps": 7496, "loss": 0.5618, "lr": 5.440000000000001e-05, "epoch": 0.21878335112059766, "percentage": 2.73, "elapsed_time": "0:40:23", "remaining_time": "23:56:19"} +{"current_steps": 210, "total_steps": 7496, "loss": 0.5645, "lr": 5.573333333333334e-05, "epoch": 0.22411953041622198, "percentage": 2.8, "elapsed_time": "0:40:51", "remaining_time": "23:37:45"} +{"current_steps": 215, "total_steps": 7496, "loss": 0.5614, "lr": 5.706666666666667e-05, "epoch": 0.2294557097118463, "percentage": 2.87, "elapsed_time": "0:41:34", "remaining_time": "23:27:58"} +{"current_steps": 220, "total_steps": 7496, "loss": 0.5601, "lr": 5.8399999999999997e-05, "epoch": 0.23479188900747064, "percentage": 2.93, "elapsed_time": "0:42:13", "remaining_time": "23:16:26"} +{"current_steps": 225, "total_steps": 7496, "loss": 0.5599, "lr": 5.973333333333334e-05, "epoch": 0.240128068303095, "percentage": 3.0, "elapsed_time": "0:42:39", "remaining_time": "22:58:28"} +{"current_steps": 230, "total_steps": 7496, "loss": 0.5607, "lr": 6.106666666666667e-05, "epoch": 0.24546424759871932, "percentage": 3.07, "elapsed_time": "0:43:19", "remaining_time": "22:48:52"} +{"current_steps": 235, "total_steps": 7496, "loss": 0.5587, "lr": 6.24e-05, "epoch": 0.2508004268943437, "percentage": 3.14, "elapsed_time": "0:43:57", "remaining_time": "22:38:20"} +{"current_steps": 240, "total_steps": 7496, "loss": 0.5571, "lr": 6.373333333333333e-05, "epoch": 0.256136606189968, "percentage": 3.2, "elapsed_time": "0:44:29", "remaining_time": "22:25:01"} +{"current_steps": 245, "total_steps": 7496, "loss": 0.5562, "lr": 6.506666666666666e-05, "epoch": 0.26147278548559233, "percentage": 3.27, "elapsed_time": "0:45:03", "remaining_time": "22:13:30"} +{"current_steps": 250, "total_steps": 7496, "loss": 0.5557, "lr": 6.64e-05, "epoch": 0.26680896478121663, "percentage": 3.34, "elapsed_time": "0:45:44", "remaining_time": "22:05:38"} +{"current_steps": 255, "total_steps": 7496, "loss": 0.5573, "lr": 6.773333333333333e-05, "epoch": 0.272145144076841, "percentage": 3.4, "elapsed_time": "0:46:16", "remaining_time": "21:53:57"} +{"current_steps": 260, "total_steps": 7496, "loss": 0.5536, "lr": 6.906666666666667e-05, "epoch": 0.27748132337246534, "percentage": 3.47, "elapsed_time": "0:46:47", "remaining_time": "21:42:16"} +{"current_steps": 265, "total_steps": 7496, "loss": 0.5542, "lr": 7.04e-05, "epoch": 0.28281750266808964, "percentage": 3.54, "elapsed_time": "0:47:29", "remaining_time": "21:35:54"} +{"current_steps": 270, "total_steps": 7496, "loss": 0.5552, "lr": 7.173333333333335e-05, "epoch": 0.288153681963714, "percentage": 3.6, "elapsed_time": "0:48:01", "remaining_time": "21:25:23"} +{"current_steps": 275, "total_steps": 7496, "loss": 0.5525, "lr": 7.306666666666668e-05, "epoch": 0.2934898612593383, "percentage": 3.67, "elapsed_time": "0:48:32", "remaining_time": "21:14:49"} +{"current_steps": 280, "total_steps": 7496, "loss": 0.5558, "lr": 7.44e-05, "epoch": 0.29882604055496265, "percentage": 3.74, "elapsed_time": "0:49:11", "remaining_time": "21:07:41"} +{"current_steps": 285, "total_steps": 7496, "loss": 0.5542, "lr": 7.573333333333334e-05, "epoch": 0.304162219850587, "percentage": 3.8, "elapsed_time": "0:49:48", "remaining_time": "21:00:03"} +{"current_steps": 290, "total_steps": 7496, "loss": 0.5512, "lr": 7.706666666666668e-05, "epoch": 0.3094983991462113, "percentage": 3.87, "elapsed_time": "0:50:20", "remaining_time": "20:50:59"} +{"current_steps": 295, "total_steps": 7496, "loss": 0.5513, "lr": 7.840000000000001e-05, "epoch": 0.31483457844183566, "percentage": 3.94, "elapsed_time": "0:50:57", "remaining_time": "20:43:49"} +{"current_steps": 300, "total_steps": 7496, "loss": 0.5508, "lr": 7.973333333333334e-05, "epoch": 0.32017075773745995, "percentage": 4.0, "elapsed_time": "0:51:34", "remaining_time": "20:37:07"} +{"current_steps": 305, "total_steps": 7496, "loss": 0.5524, "lr": 8.106666666666667e-05, "epoch": 0.3255069370330843, "percentage": 4.07, "elapsed_time": "0:52:20", "remaining_time": "20:34:12"} +{"current_steps": 310, "total_steps": 7496, "loss": 0.5456, "lr": 8.24e-05, "epoch": 0.33084311632870866, "percentage": 4.14, "elapsed_time": "0:52:55", "remaining_time": "20:26:57"} +{"current_steps": 315, "total_steps": 7496, "loss": 0.5485, "lr": 8.373333333333334e-05, "epoch": 0.33617929562433296, "percentage": 4.2, "elapsed_time": "0:53:36", "remaining_time": "20:21:55"} +{"current_steps": 320, "total_steps": 7496, "loss": 0.5457, "lr": 8.506666666666667e-05, "epoch": 0.3415154749199573, "percentage": 4.27, "elapsed_time": "0:54:09", "remaining_time": "20:14:34"} +{"current_steps": 325, "total_steps": 7496, "loss": 0.55, "lr": 8.64e-05, "epoch": 0.3468516542155816, "percentage": 4.34, "elapsed_time": "0:54:43", "remaining_time": "20:07:30"} +{"current_steps": 330, "total_steps": 7496, "loss": 0.546, "lr": 8.773333333333333e-05, "epoch": 0.35218783351120597, "percentage": 4.4, "elapsed_time": "0:55:26", "remaining_time": "20:03:53"} +{"current_steps": 335, "total_steps": 7496, "loss": 0.5464, "lr": 8.906666666666667e-05, "epoch": 0.3575240128068303, "percentage": 4.47, "elapsed_time": "0:56:02", "remaining_time": "19:57:47"} +{"current_steps": 340, "total_steps": 7496, "loss": 0.5445, "lr": 9.04e-05, "epoch": 0.3628601921024546, "percentage": 4.54, "elapsed_time": "0:56:35", "remaining_time": "19:51:09"} +{"current_steps": 345, "total_steps": 7496, "loss": 0.5452, "lr": 9.173333333333333e-05, "epoch": 0.368196371398079, "percentage": 4.6, "elapsed_time": "0:57:13", "remaining_time": "19:46:12"} +{"current_steps": 350, "total_steps": 7496, "loss": 0.5437, "lr": 9.306666666666667e-05, "epoch": 0.37353255069370334, "percentage": 4.67, "elapsed_time": "0:57:52", "remaining_time": "19:41:29"} +{"current_steps": 355, "total_steps": 7496, "loss": 0.5431, "lr": 9.44e-05, "epoch": 0.37886872998932764, "percentage": 4.74, "elapsed_time": "0:58:24", "remaining_time": "19:35:02"} +{"current_steps": 360, "total_steps": 7496, "loss": 0.5407, "lr": 9.573333333333335e-05, "epoch": 0.384204909284952, "percentage": 4.8, "elapsed_time": "0:58:59", "remaining_time": "19:29:28"} +{"current_steps": 365, "total_steps": 7496, "loss": 0.5418, "lr": 9.706666666666668e-05, "epoch": 0.3895410885805763, "percentage": 4.87, "elapsed_time": "0:59:38", "remaining_time": "19:25:17"} +{"current_steps": 370, "total_steps": 7496, "loss": 0.5414, "lr": 9.84e-05, "epoch": 0.39487726787620064, "percentage": 4.94, "elapsed_time": "1:00:11", "remaining_time": "19:19:12"} +{"current_steps": 375, "total_steps": 7496, "loss": 0.5422, "lr": 9.973333333333334e-05, "epoch": 0.400213447171825, "percentage": 5.0, "elapsed_time": "1:00:46", "remaining_time": "19:14:08"} +{"current_steps": 380, "total_steps": 7496, "loss": 0.5416, "lr": 0.00010106666666666667, "epoch": 0.4055496264674493, "percentage": 5.07, "elapsed_time": "1:01:26", "remaining_time": "19:10:31"} +{"current_steps": 385, "total_steps": 7496, "loss": 0.5404, "lr": 0.00010240000000000001, "epoch": 0.41088580576307365, "percentage": 5.14, "elapsed_time": "1:01:58", "remaining_time": "19:04:41"} +{"current_steps": 390, "total_steps": 7496, "loss": 0.5397, "lr": 0.00010373333333333335, "epoch": 0.41622198505869795, "percentage": 5.2, "elapsed_time": "1:02:34", "remaining_time": "19:00:14"} +{"current_steps": 395, "total_steps": 7496, "loss": 0.5425, "lr": 0.00010506666666666667, "epoch": 0.4215581643543223, "percentage": 5.27, "elapsed_time": "1:03:11", "remaining_time": "18:55:54"} +{"current_steps": 400, "total_steps": 7496, "loss": 0.5411, "lr": 0.00010640000000000001, "epoch": 0.42689434364994666, "percentage": 5.34, "elapsed_time": "1:03:47", "remaining_time": "18:51:33"} +{"current_steps": 405, "total_steps": 7496, "loss": 0.5399, "lr": 0.00010773333333333333, "epoch": 0.43223052294557096, "percentage": 5.4, "elapsed_time": "1:04:31", "remaining_time": "18:49:51"} +{"current_steps": 410, "total_steps": 7496, "loss": 0.5389, "lr": 0.00010906666666666667, "epoch": 0.4375667022411953, "percentage": 5.47, "elapsed_time": "1:05:10", "remaining_time": "18:46:25"} +{"current_steps": 415, "total_steps": 7496, "loss": 0.5394, "lr": 0.00011040000000000001, "epoch": 0.4429028815368196, "percentage": 5.54, "elapsed_time": "1:05:46", "remaining_time": "18:42:12"} +{"current_steps": 420, "total_steps": 7496, "loss": 0.539, "lr": 0.00011173333333333333, "epoch": 0.44823906083244397, "percentage": 5.6, "elapsed_time": "1:06:21", "remaining_time": "18:38:05"} +{"current_steps": 425, "total_steps": 7496, "loss": 0.5382, "lr": 0.00011306666666666667, "epoch": 0.4535752401280683, "percentage": 5.67, "elapsed_time": "1:06:57", "remaining_time": "18:34:09"} +{"current_steps": 430, "total_steps": 7496, "loss": 0.5345, "lr": 0.0001144, "epoch": 0.4589114194236926, "percentage": 5.74, "elapsed_time": "1:07:34", "remaining_time": "18:30:32"} +{"current_steps": 435, "total_steps": 7496, "loss": 0.5367, "lr": 0.00011573333333333333, "epoch": 0.464247598719317, "percentage": 5.8, "elapsed_time": "1:08:10", "remaining_time": "18:26:29"} +{"current_steps": 440, "total_steps": 7496, "loss": 0.5384, "lr": 0.00011706666666666668, "epoch": 0.4695837780149413, "percentage": 5.87, "elapsed_time": "1:08:45", "remaining_time": "18:22:34"} +{"current_steps": 445, "total_steps": 7496, "loss": 0.5379, "lr": 0.0001184, "epoch": 0.47491995731056563, "percentage": 5.94, "elapsed_time": "1:09:21", "remaining_time": "18:19:03"} +{"current_steps": 450, "total_steps": 7496, "loss": 0.5365, "lr": 0.00011973333333333335, "epoch": 0.48025613660619, "percentage": 6.0, "elapsed_time": "1:09:58", "remaining_time": "18:15:43"} +{"current_steps": 455, "total_steps": 7496, "loss": 0.5357, "lr": 0.00012106666666666666, "epoch": 0.4855923159018143, "percentage": 6.07, "elapsed_time": "1:10:33", "remaining_time": "18:11:45"} +{"current_steps": 460, "total_steps": 7496, "loss": 0.5338, "lr": 0.0001224, "epoch": 0.49092849519743864, "percentage": 6.14, "elapsed_time": "1:11:08", "remaining_time": "18:08:15"} +{"current_steps": 465, "total_steps": 7496, "loss": 0.5304, "lr": 0.00012373333333333335, "epoch": 0.49626467449306294, "percentage": 6.2, "elapsed_time": "1:11:48", "remaining_time": "18:05:39"} +{"current_steps": 470, "total_steps": 7496, "loss": 0.5353, "lr": 0.00012506666666666665, "epoch": 0.5016008537886874, "percentage": 6.27, "elapsed_time": "1:12:20", "remaining_time": "18:01:30"} +{"current_steps": 475, "total_steps": 7496, "loss": 0.5326, "lr": 0.0001264, "epoch": 0.5069370330843116, "percentage": 6.34, "elapsed_time": "1:12:56", "remaining_time": "17:58:07"} +{"current_steps": 480, "total_steps": 7496, "loss": 0.5344, "lr": 0.00012773333333333334, "epoch": 0.512273212379936, "percentage": 6.4, "elapsed_time": "1:13:34", "remaining_time": "17:55:25"} +{"current_steps": 485, "total_steps": 7496, "loss": 0.5342, "lr": 0.00012906666666666667, "epoch": 0.5176093916755603, "percentage": 6.47, "elapsed_time": "1:14:08", "remaining_time": "17:51:44"} +{"current_steps": 490, "total_steps": 7496, "loss": 0.5334, "lr": 0.0001304, "epoch": 0.5229455709711847, "percentage": 6.54, "elapsed_time": "1:14:41", "remaining_time": "17:48:01"} +{"current_steps": 495, "total_steps": 7496, "loss": 0.5318, "lr": 0.00013173333333333333, "epoch": 0.528281750266809, "percentage": 6.6, "elapsed_time": "1:15:19", "remaining_time": "17:45:16"} +{"current_steps": 500, "total_steps": 7496, "loss": 0.535, "lr": 0.00013306666666666668, "epoch": 0.5336179295624333, "percentage": 6.67, "elapsed_time": "1:15:56", "remaining_time": "17:42:33"} +{"current_steps": 505, "total_steps": 7496, "loss": 0.5318, "lr": 0.00013440000000000001, "epoch": 0.5389541088580576, "percentage": 6.74, "elapsed_time": "1:16:39", "remaining_time": "17:41:14"} +{"current_steps": 510, "total_steps": 7496, "loss": 0.5293, "lr": 0.00013573333333333334, "epoch": 0.544290288153682, "percentage": 6.8, "elapsed_time": "1:17:17", "remaining_time": "17:38:41"} +{"current_steps": 515, "total_steps": 7496, "loss": 0.5288, "lr": 0.00013706666666666667, "epoch": 0.5496264674493063, "percentage": 6.87, "elapsed_time": "1:17:53", "remaining_time": "17:35:54"} +{"current_steps": 520, "total_steps": 7496, "loss": 0.5318, "lr": 0.0001384, "epoch": 0.5549626467449307, "percentage": 6.94, "elapsed_time": "1:18:27", "remaining_time": "17:32:26"} +{"current_steps": 525, "total_steps": 7496, "loss": 0.5311, "lr": 0.00013973333333333333, "epoch": 0.5602988260405549, "percentage": 7.0, "elapsed_time": "1:19:01", "remaining_time": "17:29:18"} +{"current_steps": 530, "total_steps": 7496, "loss": 0.5292, "lr": 0.0001410666666666667, "epoch": 0.5656350053361793, "percentage": 7.07, "elapsed_time": "1:19:41", "remaining_time": "17:27:30"} +{"current_steps": 535, "total_steps": 7496, "loss": 0.5276, "lr": 0.0001424, "epoch": 0.5709711846318036, "percentage": 7.14, "elapsed_time": "1:20:14", "remaining_time": "17:24:02"} +{"current_steps": 540, "total_steps": 7496, "loss": 0.5289, "lr": 0.00014373333333333335, "epoch": 0.576307363927428, "percentage": 7.2, "elapsed_time": "1:20:47", "remaining_time": "17:20:49"} +{"current_steps": 545, "total_steps": 7496, "loss": 0.5251, "lr": 0.00014506666666666668, "epoch": 0.5816435432230523, "percentage": 7.27, "elapsed_time": "1:21:27", "remaining_time": "17:18:57"} +{"current_steps": 550, "total_steps": 7496, "loss": 0.5298, "lr": 0.0001464, "epoch": 0.5869797225186766, "percentage": 7.34, "elapsed_time": "1:22:00", "remaining_time": "17:15:38"} +{"current_steps": 555, "total_steps": 7496, "loss": 0.5287, "lr": 0.00014773333333333334, "epoch": 0.5923159018143009, "percentage": 7.4, "elapsed_time": "1:22:35", "remaining_time": "17:13:00"} +{"current_steps": 560, "total_steps": 7496, "loss": 0.5278, "lr": 0.00014906666666666667, "epoch": 0.5976520811099253, "percentage": 7.47, "elapsed_time": "1:23:12", "remaining_time": "17:10:35"} +{"current_steps": 565, "total_steps": 7496, "loss": 0.5264, "lr": 0.0001504, "epoch": 0.6029882604055496, "percentage": 7.54, "elapsed_time": "1:23:50", "remaining_time": "17:08:27"} +{"current_steps": 570, "total_steps": 7496, "loss": 0.5298, "lr": 0.00015173333333333335, "epoch": 0.608324439701174, "percentage": 7.6, "elapsed_time": "1:24:22", "remaining_time": "17:05:18"} +{"current_steps": 575, "total_steps": 7496, "loss": 0.5272, "lr": 0.00015306666666666666, "epoch": 0.6136606189967982, "percentage": 7.67, "elapsed_time": "1:24:58", "remaining_time": "17:02:46"} +{"current_steps": 580, "total_steps": 7496, "loss": 0.5277, "lr": 0.0001544, "epoch": 0.6189967982924226, "percentage": 7.74, "elapsed_time": "1:25:37", "remaining_time": "17:01:04"} +{"current_steps": 585, "total_steps": 7496, "loss": 0.5259, "lr": 0.00015573333333333334, "epoch": 0.624332977588047, "percentage": 7.8, "elapsed_time": "1:26:11", "remaining_time": "16:58:10"} +{"current_steps": 590, "total_steps": 7496, "loss": 0.5245, "lr": 0.00015706666666666667, "epoch": 0.6296691568836713, "percentage": 7.87, "elapsed_time": "1:26:44", "remaining_time": "16:55:19"} +{"current_steps": 595, "total_steps": 7496, "loss": 0.5261, "lr": 0.00015840000000000003, "epoch": 0.6350053361792957, "percentage": 7.94, "elapsed_time": "1:27:21", "remaining_time": "16:53:09"} +{"current_steps": 600, "total_steps": 7496, "loss": 0.5242, "lr": 0.00015973333333333333, "epoch": 0.6403415154749199, "percentage": 8.0, "elapsed_time": "1:27:55", "remaining_time": "16:50:31"} +{"current_steps": 605, "total_steps": 7496, "loss": 0.5263, "lr": 0.0001610666666666667, "epoch": 0.6456776947705443, "percentage": 8.07, "elapsed_time": "1:28:39", "remaining_time": "16:49:49"} +{"current_steps": 610, "total_steps": 7496, "loss": 0.5238, "lr": 0.00016240000000000002, "epoch": 0.6510138740661686, "percentage": 8.14, "elapsed_time": "1:29:13", "remaining_time": "16:47:16"} +{"current_steps": 615, "total_steps": 7496, "loss": 0.5227, "lr": 0.00016373333333333335, "epoch": 0.656350053361793, "percentage": 8.2, "elapsed_time": "1:29:48", "remaining_time": "16:44:45"} +{"current_steps": 620, "total_steps": 7496, "loss": 0.5244, "lr": 0.00016506666666666668, "epoch": 0.6616862326574173, "percentage": 8.27, "elapsed_time": "1:30:21", "remaining_time": "16:42:07"} +{"current_steps": 625, "total_steps": 7496, "loss": 0.5231, "lr": 0.0001664, "epoch": 0.6670224119530416, "percentage": 8.34, "elapsed_time": "1:30:56", "remaining_time": "16:39:42"} +{"current_steps": 630, "total_steps": 7496, "loss": 0.5243, "lr": 0.00016773333333333334, "epoch": 0.6723585912486659, "percentage": 8.4, "elapsed_time": "1:31:35", "remaining_time": "16:38:13"} +{"current_steps": 635, "total_steps": 7496, "loss": 0.5219, "lr": 0.0001690666666666667, "epoch": 0.6776947705442903, "percentage": 8.47, "elapsed_time": "1:32:07", "remaining_time": "16:35:22"} +{"current_steps": 640, "total_steps": 7496, "loss": 0.5226, "lr": 0.0001704, "epoch": 0.6830309498399146, "percentage": 8.54, "elapsed_time": "1:32:43", "remaining_time": "16:33:21"} +{"current_steps": 645, "total_steps": 7496, "loss": 0.5218, "lr": 0.00017173333333333335, "epoch": 0.688367129135539, "percentage": 8.6, "elapsed_time": "1:33:20", "remaining_time": "16:31:32"} +{"current_steps": 650, "total_steps": 7496, "loss": 0.5232, "lr": 0.00017306666666666665, "epoch": 0.6937033084311632, "percentage": 8.67, "elapsed_time": "1:33:56", "remaining_time": "16:29:22"} +{"current_steps": 655, "total_steps": 7496, "loss": 0.5218, "lr": 0.0001744, "epoch": 0.6990394877267876, "percentage": 8.74, "elapsed_time": "1:34:31", "remaining_time": "16:27:17"} +{"current_steps": 660, "total_steps": 7496, "loss": 0.521, "lr": 0.00017573333333333334, "epoch": 0.7043756670224119, "percentage": 8.8, "elapsed_time": "1:35:09", "remaining_time": "16:25:37"} +{"current_steps": 665, "total_steps": 7496, "loss": 0.5215, "lr": 0.00017706666666666667, "epoch": 0.7097118463180363, "percentage": 8.87, "elapsed_time": "1:35:46", "remaining_time": "16:23:45"} +{"current_steps": 670, "total_steps": 7496, "loss": 0.5203, "lr": 0.0001784, "epoch": 0.7150480256136607, "percentage": 8.94, "elapsed_time": "1:36:15", "remaining_time": "16:20:46"} +{"current_steps": 675, "total_steps": 7496, "loss": 0.5228, "lr": 0.00017973333333333333, "epoch": 0.7203842049092849, "percentage": 9.0, "elapsed_time": "1:36:53", "remaining_time": "16:19:08"} +{"current_steps": 680, "total_steps": 7496, "loss": 0.5218, "lr": 0.00018106666666666669, "epoch": 0.7257203842049093, "percentage": 9.07, "elapsed_time": "1:37:31", "remaining_time": "16:17:29"} +{"current_steps": 685, "total_steps": 7496, "loss": 0.5181, "lr": 0.00018240000000000002, "epoch": 0.7310565635005336, "percentage": 9.14, "elapsed_time": "1:38:04", "remaining_time": "16:15:05"} +{"current_steps": 690, "total_steps": 7496, "loss": 0.5199, "lr": 0.00018373333333333335, "epoch": 0.736392742796158, "percentage": 9.2, "elapsed_time": "1:38:38", "remaining_time": "16:12:58"} +{"current_steps": 695, "total_steps": 7496, "loss": 0.5186, "lr": 0.00018506666666666667, "epoch": 0.7417289220917823, "percentage": 9.27, "elapsed_time": "1:39:14", "remaining_time": "16:11:12"} +{"current_steps": 700, "total_steps": 7496, "loss": 0.5176, "lr": 0.00018640000000000003, "epoch": 0.7470651013874067, "percentage": 9.34, "elapsed_time": "1:39:48", "remaining_time": "16:08:58"} +{"current_steps": 705, "total_steps": 7496, "loss": 0.5171, "lr": 0.00018773333333333333, "epoch": 0.7524012806830309, "percentage": 9.41, "elapsed_time": "1:40:32", "remaining_time": "16:08:24"} +{"current_steps": 710, "total_steps": 7496, "loss": 0.5182, "lr": 0.0001890666666666667, "epoch": 0.7577374599786553, "percentage": 9.47, "elapsed_time": "1:41:12", "remaining_time": "16:07:21"} +{"current_steps": 715, "total_steps": 7496, "loss": 0.5134, "lr": 0.0001904, "epoch": 0.7630736392742796, "percentage": 9.54, "elapsed_time": "1:41:47", "remaining_time": "16:05:21"} +{"current_steps": 720, "total_steps": 7496, "loss": 0.5173, "lr": 0.00019173333333333335, "epoch": 0.768409818569904, "percentage": 9.61, "elapsed_time": "1:42:21", "remaining_time": "16:03:16"} +{"current_steps": 725, "total_steps": 7496, "loss": 0.516, "lr": 0.00019306666666666668, "epoch": 0.7737459978655283, "percentage": 9.67, "elapsed_time": "1:42:56", "remaining_time": "16:01:28"} +{"current_steps": 730, "total_steps": 7496, "loss": 0.5167, "lr": 0.0001944, "epoch": 0.7790821771611526, "percentage": 9.74, "elapsed_time": "1:43:33", "remaining_time": "15:59:45"} +{"current_steps": 735, "total_steps": 7496, "loss": 0.5156, "lr": 0.00019573333333333334, "epoch": 0.7844183564567769, "percentage": 9.81, "elapsed_time": "1:44:06", "remaining_time": "15:57:35"} +{"current_steps": 740, "total_steps": 7496, "loss": 0.5153, "lr": 0.00019706666666666667, "epoch": 0.7897545357524013, "percentage": 9.87, "elapsed_time": "1:44:40", "remaining_time": "15:55:42"} +{"current_steps": 745, "total_steps": 7496, "loss": 0.514, "lr": 0.0001984, "epoch": 0.7950907150480256, "percentage": 9.94, "elapsed_time": "1:45:17", "remaining_time": "15:54:03"} +{"current_steps": 750, "total_steps": 7496, "loss": 0.5153, "lr": 0.00019973333333333335, "epoch": 0.80042689434365, "percentage": 10.01, "elapsed_time": "1:45:52", "remaining_time": "15:52:18"} +{"current_steps": 755, "total_steps": 7496, "loss": 0.5113, "lr": 0.00019999982650101714, "epoch": 0.8057630736392742, "percentage": 10.07, "elapsed_time": "1:46:26", "remaining_time": "15:50:24"} +{"current_steps": 760, "total_steps": 7496, "loss": 0.5131, "lr": 0.00019999912166243107, "epoch": 0.8110992529348986, "percentage": 10.14, "elapsed_time": "1:47:02", "remaining_time": "15:48:45"} +{"current_steps": 765, "total_steps": 7496, "loss": 0.5116, "lr": 0.00019999787464437395, "epoch": 0.816435432230523, "percentage": 10.21, "elapsed_time": "1:47:37", "remaining_time": "15:47:01"} +{"current_steps": 770, "total_steps": 7496, "loss": 0.5189, "lr": 0.0001999960854536069, "epoch": 0.8217716115261473, "percentage": 10.27, "elapsed_time": "1:48:13", "remaining_time": "15:45:23"} +{"current_steps": 775, "total_steps": 7496, "loss": 0.5155, "lr": 0.00019999375409983057, "epoch": 0.8271077908217717, "percentage": 10.34, "elapsed_time": "1:48:48", "remaining_time": "15:43:37"} +{"current_steps": 780, "total_steps": 7496, "loss": 0.5129, "lr": 0.00019999088059568536, "epoch": 0.8324439701173959, "percentage": 10.41, "elapsed_time": "1:49:22", "remaining_time": "15:41:41"} +{"current_steps": 785, "total_steps": 7496, "loss": 0.5128, "lr": 0.0001999874649567508, "epoch": 0.8377801494130203, "percentage": 10.47, "elapsed_time": "1:49:55", "remaining_time": "15:39:47"} +{"current_steps": 790, "total_steps": 7496, "loss": 0.513, "lr": 0.00019998350720154605, "epoch": 0.8431163287086446, "percentage": 10.54, "elapsed_time": "1:50:33", "remaining_time": "15:38:32"} +{"current_steps": 795, "total_steps": 7496, "loss": 0.5111, "lr": 0.00019997900735152936, "epoch": 0.848452508004269, "percentage": 10.61, "elapsed_time": "1:51:08", "remaining_time": "15:36:48"} +{"current_steps": 800, "total_steps": 7496, "loss": 0.514, "lr": 0.0001999739654310982, "epoch": 0.8537886872998933, "percentage": 10.67, "elapsed_time": "1:51:43", "remaining_time": "15:35:07"} +{"current_steps": 805, "total_steps": 7496, "loss": 0.5115, "lr": 0.00019996838146758915, "epoch": 0.8591248665955176, "percentage": 10.74, "elapsed_time": "1:52:34", "remaining_time": "15:35:38"} +{"current_steps": 810, "total_steps": 7496, "loss": 0.5119, "lr": 0.00019996225549127753, "epoch": 0.8644610458911419, "percentage": 10.81, "elapsed_time": "1:53:09", "remaining_time": "15:34:05"} +{"current_steps": 815, "total_steps": 7496, "loss": 0.5111, "lr": 0.0001999555875353774, "epoch": 0.8697972251867663, "percentage": 10.87, "elapsed_time": "1:53:44", "remaining_time": "15:32:26"} +{"current_steps": 820, "total_steps": 7496, "loss": 0.5101, "lr": 0.0001999483776360414, "epoch": 0.8751334044823906, "percentage": 10.94, "elapsed_time": "1:54:21", "remaining_time": "15:31:02"} +{"current_steps": 825, "total_steps": 7496, "loss": 0.5078, "lr": 0.00019994062583236045, "epoch": 0.880469583778015, "percentage": 11.01, "elapsed_time": "1:54:56", "remaining_time": "15:29:28"} +{"current_steps": 830, "total_steps": 7496, "loss": 0.5115, "lr": 0.00019993233216636358, "epoch": 0.8858057630736392, "percentage": 11.07, "elapsed_time": "1:55:32", "remaining_time": "15:27:54"} +{"current_steps": 835, "total_steps": 7496, "loss": 0.5083, "lr": 0.0001999234966830178, "epoch": 0.8911419423692636, "percentage": 11.14, "elapsed_time": "1:56:08", "remaining_time": "15:26:27"} +{"current_steps": 840, "total_steps": 7496, "loss": 0.5114, "lr": 0.00019991411943022768, "epoch": 0.8964781216648879, "percentage": 11.21, "elapsed_time": "1:56:46", "remaining_time": "15:25:20"} +{"current_steps": 845, "total_steps": 7496, "loss": 0.5093, "lr": 0.00019990420045883518, "epoch": 0.9018143009605123, "percentage": 11.27, "elapsed_time": "1:57:21", "remaining_time": "15:23:45"} +{"current_steps": 850, "total_steps": 7496, "loss": 0.5075, "lr": 0.00019989373982261946, "epoch": 0.9071504802561366, "percentage": 11.34, "elapsed_time": "1:57:58", "remaining_time": "15:22:25"} +{"current_steps": 855, "total_steps": 7496, "loss": 0.508, "lr": 0.0001998827375782964, "epoch": 0.9124866595517609, "percentage": 11.41, "elapsed_time": "1:58:35", "remaining_time": "15:21:10"} +{"current_steps": 860, "total_steps": 7496, "loss": 0.5064, "lr": 0.0001998711937855184, "epoch": 0.9178228388473852, "percentage": 11.47, "elapsed_time": "1:59:10", "remaining_time": "15:19:37"} +{"current_steps": 865, "total_steps": 7496, "loss": 0.5087, "lr": 0.00019985910850687413, "epoch": 0.9231590181430096, "percentage": 11.54, "elapsed_time": "1:59:46", "remaining_time": "15:18:11"} +{"current_steps": 870, "total_steps": 7496, "loss": 0.5043, "lr": 0.00019984648180788804, "epoch": 0.928495197438634, "percentage": 11.61, "elapsed_time": "2:00:21", "remaining_time": "15:16:40"} +{"current_steps": 875, "total_steps": 7496, "loss": 0.5032, "lr": 0.00019983331375702016, "epoch": 0.9338313767342583, "percentage": 11.67, "elapsed_time": "2:00:58", "remaining_time": "15:15:21"} +{"current_steps": 880, "total_steps": 7496, "loss": 0.5054, "lr": 0.00019981960442566553, "epoch": 0.9391675560298826, "percentage": 11.74, "elapsed_time": "2:01:34", "remaining_time": "15:14:02"} +{"current_steps": 885, "total_steps": 7496, "loss": 0.5052, "lr": 0.000199805353888154, "epoch": 0.9445037353255069, "percentage": 11.81, "elapsed_time": "2:02:11", "remaining_time": "15:12:47"} +{"current_steps": 890, "total_steps": 7496, "loss": 0.5058, "lr": 0.00019979056222174972, "epoch": 0.9498399146211313, "percentage": 11.87, "elapsed_time": "2:02:46", "remaining_time": "15:11:20"} +{"current_steps": 895, "total_steps": 7496, "loss": 0.5051, "lr": 0.0001997752295066508, "epoch": 0.9551760939167556, "percentage": 11.94, "elapsed_time": "2:03:24", "remaining_time": "15:10:08"} +{"current_steps": 900, "total_steps": 7496, "loss": 0.5037, "lr": 0.0001997593558259888, "epoch": 0.96051227321238, "percentage": 12.01, "elapsed_time": "2:03:58", "remaining_time": "15:08:38"} +{"current_steps": 905, "total_steps": 7496, "loss": 0.5067, "lr": 0.00019974294126582826, "epoch": 0.9658484525080042, "percentage": 12.07, "elapsed_time": "2:04:49", "remaining_time": "15:09:06"} +{"current_steps": 910, "total_steps": 7496, "loss": 0.5069, "lr": 0.00019972598591516642, "epoch": 0.9711846318036286, "percentage": 12.14, "elapsed_time": "2:05:25", "remaining_time": "15:07:43"} +{"current_steps": 915, "total_steps": 7496, "loss": 0.5071, "lr": 0.00019970848986593248, "epoch": 0.9765208110992529, "percentage": 12.21, "elapsed_time": "2:05:58", "remaining_time": "15:06:06"} +{"current_steps": 920, "total_steps": 7496, "loss": 0.5071, "lr": 0.00019969045321298725, "epoch": 0.9818569903948773, "percentage": 12.27, "elapsed_time": "2:06:40", "remaining_time": "15:05:23"} +{"current_steps": 905, "total_steps": 7496, "loss": 0.5068, "lr": 0.00019974294126582826, "epoch": 0.9658484525080042, "percentage": 12.07, "elapsed_time": "0:07:13", "remaining_time": "0:52:36"} +{"current_steps": 910, "total_steps": 7496, "loss": 0.5069, "lr": 0.00019972598591516642, "epoch": 0.9711846318036286, "percentage": 12.14, "elapsed_time": "0:07:37", "remaining_time": "0:55:14"} +{"current_steps": 915, "total_steps": 7496, "loss": 0.5072, "lr": 0.00019970848986593248, "epoch": 0.9765208110992529, "percentage": 12.21, "elapsed_time": "0:08:02", "remaining_time": "0:57:48"} +{"current_steps": 920, "total_steps": 7496, "loss": 0.5072, "lr": 0.00019969045321298725, "epoch": 0.9818569903948773, "percentage": 12.27, "elapsed_time": "0:08:58", "remaining_time": "1:04:12"} +{"current_steps": 925, "total_steps": 7496, "loss": 0.5035, "lr": 0.00019967187605412262, "epoch": 0.9871931696905016, "percentage": 12.34, "elapsed_time": "0:09:23", "remaining_time": "1:06:44"} +{"current_steps": 930, "total_steps": 7496, "loss": 0.5036, "lr": 0.00019965275849006102, "epoch": 0.9925293489861259, "percentage": 12.41, "elapsed_time": "0:09:47", "remaining_time": "1:09:07"} +{"current_steps": 935, "total_steps": 7496, "loss": 0.5022, "lr": 0.00019963310062445487, "epoch": 0.9978655282817502, "percentage": 12.47, "elapsed_time": "0:10:40", "remaining_time": "1:14:57"} +{"current_steps": 940, "total_steps": 7496, "loss": 0.501, "lr": 0.00019961290256388604, "epoch": 1.0032017075773747, "percentage": 12.54, "elapsed_time": "0:11:06", "remaining_time": "1:17:31"} +{"current_steps": 945, "total_steps": 7496, "loss": 0.4982, "lr": 0.00019959216441786524, "epoch": 1.0085378868729988, "percentage": 12.61, "elapsed_time": "0:11:30", "remaining_time": "1:19:50"} +{"current_steps": 950, "total_steps": 7496, "loss": 0.4987, "lr": 0.00019957088629883135, "epoch": 1.0138740661686232, "percentage": 12.67, "elapsed_time": "0:12:31", "remaining_time": "1:26:17"} +{"current_steps": 955, "total_steps": 7496, "loss": 0.4972, "lr": 0.00019954906832215103, "epoch": 1.0192102454642475, "percentage": 12.74, "elapsed_time": "0:13:06", "remaining_time": "1:29:46"} +{"current_steps": 960, "total_steps": 7496, "loss": 0.5008, "lr": 0.00019952671060611792, "epoch": 1.024546424759872, "percentage": 12.81, "elapsed_time": "0:13:29", "remaining_time": "1:31:52"} +{"current_steps": 965, "total_steps": 7496, "loss": 0.5004, "lr": 0.000199503813271952, "epoch": 1.0298826040554963, "percentage": 12.87, "elapsed_time": "0:14:09", "remaining_time": "1:35:48"} +{"current_steps": 970, "total_steps": 7496, "loss": 0.4979, "lr": 0.00019948037644379905, "epoch": 1.0352187833511206, "percentage": 12.94, "elapsed_time": "0:14:51", "remaining_time": "1:39:56"} +{"current_steps": 975, "total_steps": 7496, "loss": 0.4993, "lr": 0.0001994564002487298, "epoch": 1.040554962646745, "percentage": 13.01, "elapsed_time": "0:15:15", "remaining_time": "1:42:01"} +{"current_steps": 980, "total_steps": 7496, "loss": 0.496, "lr": 0.00019943188481673946, "epoch": 1.0458911419423693, "percentage": 13.07, "elapsed_time": "0:15:45", "remaining_time": "1:44:47"} +{"current_steps": 985, "total_steps": 7496, "loss": 0.4974, "lr": 0.00019940683028074675, "epoch": 1.0512273212379937, "percentage": 13.14, "elapsed_time": "0:16:39", "remaining_time": "1:50:04"} +{"current_steps": 990, "total_steps": 7496, "loss": 0.4973, "lr": 0.00019938123677659352, "epoch": 1.056563500533618, "percentage": 13.21, "elapsed_time": "0:17:05", "remaining_time": "1:52:17"} +{"current_steps": 995, "total_steps": 7496, "loss": 0.497, "lr": 0.00019935510444304364, "epoch": 1.0618996798292422, "percentage": 13.27, "elapsed_time": "0:17:28", "remaining_time": "1:54:11"} +{"current_steps": 1000, "total_steps": 7496, "loss": 0.4966, "lr": 0.00019932843342178254, "epoch": 1.0672358591248665, "percentage": 13.34, "elapsed_time": "0:18:44", "remaining_time": "2:01:43"} +{"current_steps": 1005, "total_steps": 7496, "loss": 0.5005, "lr": 0.00019930122385741625, "epoch": 1.0725720384204909, "percentage": 13.41, "elapsed_time": "0:19:29", "remaining_time": "2:05:54"} +{"current_steps": 1010, "total_steps": 7496, "loss": 0.4981, "lr": 0.00019927347589747082, "epoch": 1.0779082177161152, "percentage": 13.47, "elapsed_time": "0:19:52", "remaining_time": "2:07:40"} +{"current_steps": 1015, "total_steps": 7496, "loss": 0.4957, "lr": 0.0001992451896923912, "epoch": 1.0832443970117396, "percentage": 13.54, "elapsed_time": "0:20:36", "remaining_time": "2:11:36"} +{"current_steps": 1020, "total_steps": 7496, "loss": 0.4977, "lr": 0.00019921636539554084, "epoch": 1.088580576307364, "percentage": 13.61, "elapsed_time": "0:21:56", "remaining_time": "2:19:15"} +{"current_steps": 1025, "total_steps": 7496, "loss": 0.4958, "lr": 0.00019918700316320044, "epoch": 1.0939167556029883, "percentage": 13.67, "elapsed_time": "0:22:20", "remaining_time": "2:21:02"} +{"current_steps": 1030, "total_steps": 7496, "loss": 0.4927, "lr": 0.00019915710315456748, "epoch": 1.0992529348986126, "percentage": 13.74, "elapsed_time": "0:22:55", "remaining_time": "2:23:57"} +{"current_steps": 1035, "total_steps": 7496, "loss": 0.4944, "lr": 0.00019912666553175502, "epoch": 1.104589114194237, "percentage": 13.81, "elapsed_time": "0:23:45", "remaining_time": "2:28:16"} +{"current_steps": 1040, "total_steps": 7496, "loss": 0.4937, "lr": 0.0001990956904597911, "epoch": 1.1099252934898614, "percentage": 13.87, "elapsed_time": "0:24:13", "remaining_time": "2:30:25"} +{"current_steps": 1045, "total_steps": 7496, "loss": 0.4941, "lr": 0.0001990641781066177, "epoch": 1.1152614727854857, "percentage": 13.94, "elapsed_time": "0:24:43", "remaining_time": "2:32:36"} +{"current_steps": 1050, "total_steps": 7496, "loss": 0.4937, "lr": 0.00019903212864308975, "epoch": 1.1205976520811098, "percentage": 14.01, "elapsed_time": "0:25:32", "remaining_time": "2:36:49"} +{"current_steps": 1055, "total_steps": 7496, "loss": 0.4953, "lr": 0.00019899954224297442, "epoch": 1.1259338313767342, "percentage": 14.07, "elapsed_time": "0:26:07", "remaining_time": "2:39:30"} +{"current_steps": 1060, "total_steps": 7496, "loss": 0.4917, "lr": 0.00019896641908295006, "epoch": 1.1312700106723586, "percentage": 14.14, "elapsed_time": "0:26:33", "remaining_time": "2:41:17"} +{"current_steps": 1065, "total_steps": 7496, "loss": 0.4943, "lr": 0.00019893275934260523, "epoch": 1.136606189967983, "percentage": 14.21, "elapsed_time": "0:27:19", "remaining_time": "2:44:58"} +{"current_steps": 1070, "total_steps": 7496, "loss": 0.4932, "lr": 0.00019889856320443768, "epoch": 1.1419423692636073, "percentage": 14.27, "elapsed_time": "0:27:59", "remaining_time": "2:48:07"} +{"current_steps": 1075, "total_steps": 7496, "loss": 0.4949, "lr": 0.00019886383085385351, "epoch": 1.1472785485592316, "percentage": 14.34, "elapsed_time": "0:28:25", "remaining_time": "2:49:49"} +{"current_steps": 1080, "total_steps": 7496, "loss": 0.4939, "lr": 0.00019882856247916607, "epoch": 1.152614727854856, "percentage": 14.41, "elapsed_time": "0:29:03", "remaining_time": "2:52:36"} +{"current_steps": 1085, "total_steps": 7496, "loss": 0.4919, "lr": 0.00019879275827159486, "epoch": 1.1579509071504803, "percentage": 14.47, "elapsed_time": "0:29:47", "remaining_time": "2:56:03"} +{"current_steps": 1090, "total_steps": 7496, "loss": 0.4913, "lr": 0.00019875641842526473, "epoch": 1.1632870864461047, "percentage": 14.54, "elapsed_time": "0:30:19", "remaining_time": "2:58:11"} +{"current_steps": 1095, "total_steps": 7496, "loss": 0.4903, "lr": 0.00019871954313720454, "epoch": 1.1686232657417288, "percentage": 14.61, "elapsed_time": "0:30:50", "remaining_time": "3:00:20"} +{"current_steps": 1100, "total_steps": 7496, "loss": 0.4921, "lr": 0.00019868213260734635, "epoch": 1.1739594450373532, "percentage": 14.67, "elapsed_time": "0:31:37", "remaining_time": "3:03:51"} +{"current_steps": 1105, "total_steps": 7496, "loss": 0.4919, "lr": 0.00019864418703852406, "epoch": 1.1792956243329775, "percentage": 14.74, "elapsed_time": "0:32:24", "remaining_time": "3:07:23"} +{"current_steps": 1110, "total_steps": 7496, "loss": 0.4909, "lr": 0.00019860570663647264, "epoch": 1.1846318036286019, "percentage": 14.81, "elapsed_time": "0:32:55", "remaining_time": "3:09:23"} +{"current_steps": 1115, "total_steps": 7496, "loss": 0.4923, "lr": 0.00019856669160982674, "epoch": 1.1899679829242262, "percentage": 14.87, "elapsed_time": "0:33:37", "remaining_time": "3:12:26"} +{"current_steps": 1105, "total_steps": 7496, "loss": 0.492, "lr": 0.00019864418703852406, "epoch": 1.1792956243329775, "percentage": 14.74, "elapsed_time": "0:21:47", "remaining_time": "2:06:03"} +{"current_steps": 1110, "total_steps": 7496, "loss": 0.4931, "lr": 0.00019860570663647264, "epoch": 1.1846318036286019, "percentage": 14.81, "elapsed_time": "0:22:13", "remaining_time": "2:07:54"} +{"current_steps": 1115, "total_steps": 7496, "loss": 0.4905, "lr": 0.00019856669160982674, "epoch": 1.1899679829242262, "percentage": 14.87, "elapsed_time": "0:22:41", "remaining_time": "2:09:53"} +{"current_steps": 1120, "total_steps": 7496, "loss": 0.4908, "lr": 0.00019852714217011966, "epoch": 1.1953041622198506, "percentage": 14.94, "elapsed_time": "0:23:42", "remaining_time": "2:14:57"} +{"current_steps": 1125, "total_steps": 7496, "loss": 0.4942, "lr": 0.0001984870585317823, "epoch": 1.200640341515475, "percentage": 15.01, "elapsed_time": "0:25:12", "remaining_time": "2:22:48"} +{"current_steps": 1130, "total_steps": 7496, "loss": 0.4921, "lr": 0.00019844644091214172, "epoch": 1.2059765208110993, "percentage": 15.07, "elapsed_time": "0:25:37", "remaining_time": "2:24:23"} +{"current_steps": 1135, "total_steps": 7496, "loss": 0.4922, "lr": 0.00019840528953142032, "epoch": 1.2113127001067236, "percentage": 15.14, "elapsed_time": "0:26:29", "remaining_time": "2:28:30"} +{"current_steps": 1140, "total_steps": 7496, "loss": 0.4924, "lr": 0.0001983636046127344, "epoch": 1.216648879402348, "percentage": 15.21, "elapsed_time": "0:26:53", "remaining_time": "2:29:54"} +{"current_steps": 1145, "total_steps": 7496, "loss": 0.4897, "lr": 0.00019832138638209296, "epoch": 1.2219850586979724, "percentage": 15.27, "elapsed_time": "0:27:18", "remaining_time": "2:31:29"} +{"current_steps": 1150, "total_steps": 7496, "loss": 0.4926, "lr": 0.00019827863506839666, "epoch": 1.2273212379935965, "percentage": 15.34, "elapsed_time": "0:28:08", "remaining_time": "2:35:18"} +{"current_steps": 1155, "total_steps": 7496, "loss": 0.492, "lr": 0.00019823535090343636, "epoch": 1.2326574172892208, "percentage": 15.41, "elapsed_time": "0:30:51", "remaining_time": "2:49:26"} +{"current_steps": 1160, "total_steps": 7496, "loss": 0.4888, "lr": 0.00019819153412189194, "epoch": 1.2379935965848452, "percentage": 15.47, "elapsed_time": "0:33:09", "remaining_time": "3:01:06"} +{"current_steps": 1105, "total_steps": 7496, "loss": 0.4921, "lr": 0.00019864418703852406, "epoch": 1.1792956243329775, "percentage": 14.74, "elapsed_time": "0:19:20", "remaining_time": "1:51:50"} +{"current_steps": 1110, "total_steps": 7496, "loss": 0.4931, "lr": 0.00019860570663647264, "epoch": 1.1846318036286019, "percentage": 14.81, "elapsed_time": "0:19:46", "remaining_time": "1:53:46"} +{"current_steps": 1115, "total_steps": 7496, "loss": 0.4902, "lr": 0.00019856669160982674, "epoch": 1.1899679829242262, "percentage": 14.87, "elapsed_time": "0:20:14", "remaining_time": "1:55:53"} +{"current_steps": 1120, "total_steps": 7496, "loss": 0.4901, "lr": 0.00019852714217011966, "epoch": 1.1953041622198506, "percentage": 14.94, "elapsed_time": "0:21:43", "remaining_time": "2:03:41"} +{"current_steps": 1125, "total_steps": 7496, "loss": 0.4946, "lr": 0.0001984870585317823, "epoch": 1.200640341515475, "percentage": 15.01, "elapsed_time": "0:22:07", "remaining_time": "2:05:17"} +{"current_steps": 1130, "total_steps": 7496, "loss": 0.4918, "lr": 0.00019844644091214172, "epoch": 1.2059765208110993, "percentage": 15.07, "elapsed_time": "0:22:31", "remaining_time": "2:06:53"} +{"current_steps": 1135, "total_steps": 7496, "loss": 0.4919, "lr": 0.00019840528953142032, "epoch": 1.2113127001067236, "percentage": 15.14, "elapsed_time": "0:23:20", "remaining_time": "2:10:49"} +{"current_steps": 1140, "total_steps": 7496, "loss": 0.4924, "lr": 0.0001983636046127344, "epoch": 1.216648879402348, "percentage": 15.21, "elapsed_time": "0:23:44", "remaining_time": "2:12:20"} +{"current_steps": 1145, "total_steps": 7496, "loss": 0.4895, "lr": 0.00019832138638209296, "epoch": 1.2219850586979724, "percentage": 15.27, "elapsed_time": "0:24:08", "remaining_time": "2:13:56"} +{"current_steps": 1150, "total_steps": 7496, "loss": 0.4918, "lr": 0.00019827863506839666, "epoch": 1.2273212379935965, "percentage": 15.34, "elapsed_time": "0:25:00", "remaining_time": "2:18:01"} +{"current_steps": 1155, "total_steps": 7496, "loss": 0.4906, "lr": 0.00019823535090343636, "epoch": 1.2326574172892208, "percentage": 15.41, "elapsed_time": "0:25:30", "remaining_time": "2:20:04"} +{"current_steps": 1160, "total_steps": 7496, "loss": 0.4877, "lr": 0.00019819153412189194, "epoch": 1.2379935965848452, "percentage": 15.47, "elapsed_time": "0:25:54", "remaining_time": "2:21:32"} +{"current_steps": 1165, "total_steps": 7496, "loss": 0.4894, "lr": 0.00019814718496133108, "epoch": 1.2433297758804696, "percentage": 15.54, "elapsed_time": "0:26:34", "remaining_time": "2:24:27"} +{"current_steps": 1170, "total_steps": 7496, "loss": 0.4889, "lr": 0.00019810230366220803, "epoch": 1.248665955176094, "percentage": 15.61, "elapsed_time": "0:27:19", "remaining_time": "2:27:43"} +{"current_steps": 1175, "total_steps": 7496, "loss": 0.4893, "lr": 0.00019805689046786202, "epoch": 1.2540021344717183, "percentage": 15.68, "elapsed_time": "0:27:43", "remaining_time": "2:29:07"} +{"current_steps": 1180, "total_steps": 7496, "loss": 0.4904, "lr": 0.00019801094562451628, "epoch": 1.2593383137673426, "percentage": 15.74, "elapsed_time": "0:28:12", "remaining_time": "2:30:56"} +{"current_steps": 1185, "total_steps": 7496, "loss": 0.4893, "lr": 0.00019796446938127646, "epoch": 1.264674493062967, "percentage": 15.81, "elapsed_time": "0:29:06", "remaining_time": "2:35:03"} +{"current_steps": 1190, "total_steps": 7496, "loss": 0.49, "lr": 0.00019791746199012948, "epoch": 1.2700106723585913, "percentage": 15.88, "elapsed_time": "0:29:33", "remaining_time": "2:36:38"} +{"current_steps": 1195, "total_steps": 7496, "loss": 0.4868, "lr": 0.00019786992370594196, "epoch": 1.2753468516542155, "percentage": 15.94, "elapsed_time": "0:29:58", "remaining_time": "2:38:01"} +{"current_steps": 1200, "total_steps": 7496, "loss": 0.4884, "lr": 0.00019782185478645902, "epoch": 1.2806830309498398, "percentage": 16.01, "elapsed_time": "0:30:52", "remaining_time": "2:41:58"} +{"current_steps": 1205, "total_steps": 7496, "loss": 0.4878, "lr": 0.00019777325549230273, "epoch": 1.2860192102454642, "percentage": 16.08, "elapsed_time": "0:31:38", "remaining_time": "2:45:11"} +{"current_steps": 1210, "total_steps": 7496, "loss": 0.4917, "lr": 0.00019772412608697079, "epoch": 1.2913553895410885, "percentage": 16.14, "elapsed_time": "0:32:02", "remaining_time": "2:46:27"} +{"current_steps": 1215, "total_steps": 7496, "loss": 0.4907, "lr": 0.00019767446683683514, "epoch": 1.2966915688367129, "percentage": 16.21, "elapsed_time": "0:32:46", "remaining_time": "2:49:25"} +{"current_steps": 1220, "total_steps": 7496, "loss": 0.4883, "lr": 0.00019762427801114037, "epoch": 1.3020277481323372, "percentage": 16.28, "elapsed_time": "0:33:29", "remaining_time": "2:52:18"} +{"current_steps": 1225, "total_steps": 7496, "loss": 0.4874, "lr": 0.00019757355988200237, "epoch": 1.3073639274279616, "percentage": 16.34, "elapsed_time": "0:33:54", "remaining_time": "2:53:36"} +{"current_steps": 1230, "total_steps": 7496, "loss": 0.4863, "lr": 0.00019752231272440683, "epoch": 1.312700106723586, "percentage": 16.41, "elapsed_time": "0:34:59", "remaining_time": "2:58:17"} +{"current_steps": 1235, "total_steps": 7496, "loss": 0.4868, "lr": 0.00019747053681620785, "epoch": 1.3180362860192103, "percentage": 16.48, "elapsed_time": "0:36:24", "remaining_time": "3:04:35"} +{"current_steps": 1240, "total_steps": 7496, "loss": 0.4885, "lr": 0.00019741823243812613, "epoch": 1.3233724653148347, "percentage": 16.54, "elapsed_time": "0:36:53", "remaining_time": "3:06:06"} +{"current_steps": 1245, "total_steps": 7496, "loss": 0.4872, "lr": 0.0001973653998737478, "epoch": 1.328708644610459, "percentage": 16.61, "elapsed_time": "0:37:24", "remaining_time": "3:07:48"} +{"current_steps": 1250, "total_steps": 7496, "loss": 0.4843, "lr": 0.0001973120394095227, "epoch": 1.3340448239060834, "percentage": 16.68, "elapsed_time": "0:38:13", "remaining_time": "3:11:02"} +{"current_steps": 1255, "total_steps": 7496, "loss": 0.4849, "lr": 0.00019725815133476288, "epoch": 1.3393810032017075, "percentage": 16.74, "elapsed_time": "0:38:46", "remaining_time": "3:12:49"} +{"current_steps": 1260, "total_steps": 7496, "loss": 0.4854, "lr": 0.00019720373594164095, "epoch": 1.3447171824973319, "percentage": 16.81, "elapsed_time": "0:39:12", "remaining_time": "3:14:01"} +{"current_steps": 1265, "total_steps": 7496, "loss": 0.4865, "lr": 0.00019714879352518858, "epoch": 1.3500533617929562, "percentage": 16.88, "elapsed_time": "0:39:59", "remaining_time": "3:17:00"} +{"current_steps": 1270, "total_steps": 7496, "loss": 0.4859, "lr": 0.0001970933243832949, "epoch": 1.3553895410885806, "percentage": 16.94, "elapsed_time": "0:40:39", "remaining_time": "3:19:19"} +{"current_steps": 1275, "total_steps": 7496, "loss": 0.4847, "lr": 0.00019703732881670488, "epoch": 1.360725720384205, "percentage": 17.01, "elapsed_time": "0:41:05", "remaining_time": "3:20:29"} +{"current_steps": 1280, "total_steps": 7496, "loss": 0.4846, "lr": 0.00019698080712901765, "epoch": 1.3660618996798293, "percentage": 17.08, "elapsed_time": "0:41:44", "remaining_time": "3:22:41"} +{"current_steps": 1285, "total_steps": 7496, "loss": 0.4835, "lr": 0.00019692375962668485, "epoch": 1.3713980789754536, "percentage": 17.14, "elapsed_time": "0:42:29", "remaining_time": "3:25:20"} +{"current_steps": 1290, "total_steps": 7496, "loss": 0.4848, "lr": 0.0001968661866190091, "epoch": 1.376734258271078, "percentage": 17.21, "elapsed_time": "0:42:57", "remaining_time": "3:26:41"} +{"current_steps": 1295, "total_steps": 7496, "loss": 0.4823, "lr": 0.00019680808841814206, "epoch": 1.3820704375667021, "percentage": 17.28, "elapsed_time": "0:43:33", "remaining_time": "3:28:33"} +{"current_steps": 1300, "total_steps": 7496, "loss": 0.482, "lr": 0.00019674946533908315, "epoch": 1.3874066168623265, "percentage": 17.34, "elapsed_time": "0:44:16", "remaining_time": "3:31:00"} +{"current_steps": 1305, "total_steps": 7496, "loss": 0.4819, "lr": 0.0001966903176996774, "epoch": 1.3927427961579508, "percentage": 17.41, "elapsed_time": "0:45:02", "remaining_time": "3:33:42"} +{"current_steps": 1310, "total_steps": 7496, "loss": 0.4828, "lr": 0.00019663064582061397, "epoch": 1.3980789754535752, "percentage": 17.48, "elapsed_time": "0:45:33", "remaining_time": "3:35:08"} +{"current_steps": 1315, "total_steps": 7496, "loss": 0.4829, "lr": 0.00019657045002542442, "epoch": 1.4034151547491995, "percentage": 17.54, "elapsed_time": "0:46:17", "remaining_time": "3:37:37"} +{"current_steps": 1320, "total_steps": 7496, "loss": 0.4808, "lr": 0.00019650973064048083, "epoch": 1.4087513340448239, "percentage": 17.61, "elapsed_time": "0:46:56", "remaining_time": "3:39:36"} +{"current_steps": 1325, "total_steps": 7496, "loss": 0.4826, "lr": 0.00019644848799499413, "epoch": 1.4140875133404482, "percentage": 17.68, "elapsed_time": "0:47:25", "remaining_time": "3:40:51"} +{"current_steps": 1330, "total_steps": 7496, "loss": 0.4823, "lr": 0.0001963867224210123, "epoch": 1.4194236926360726, "percentage": 17.74, "elapsed_time": "0:48:08", "remaining_time": "3:43:10"} +{"current_steps": 1335, "total_steps": 7496, "loss": 0.4831, "lr": 0.00019632443425341854, "epoch": 1.424759871931697, "percentage": 17.81, "elapsed_time": "0:48:49", "remaining_time": "3:45:19"} +{"current_steps": 1340, "total_steps": 7496, "loss": 0.4806, "lr": 0.0001962616238299295, "epoch": 1.4300960512273213, "percentage": 17.88, "elapsed_time": "0:49:20", "remaining_time": "3:46:42"} +{"current_steps": 1345, "total_steps": 7496, "loss": 0.4803, "lr": 0.00019619829149109336, "epoch": 1.4354322305229457, "percentage": 17.94, "elapsed_time": "0:49:59", "remaining_time": "3:48:39"} +{"current_steps": 1350, "total_steps": 7496, "loss": 0.4804, "lr": 0.0001961344375802881, "epoch": 1.44076840981857, "percentage": 18.01, "elapsed_time": "0:50:42", "remaining_time": "3:50:50"} +{"current_steps": 1355, "total_steps": 7496, "loss": 0.4798, "lr": 0.0001960700624437195, "epoch": 1.4461045891141944, "percentage": 18.08, "elapsed_time": "0:51:17", "remaining_time": "3:52:27"} +{"current_steps": 1360, "total_steps": 7496, "loss": 0.4808, "lr": 0.00019600516643041943, "epoch": 1.4514407684098185, "percentage": 18.14, "elapsed_time": "0:51:51", "remaining_time": "3:53:58"} +{"current_steps": 1365, "total_steps": 7496, "loss": 0.4804, "lr": 0.0001959397498922439, "epoch": 1.4567769477054429, "percentage": 18.21, "elapsed_time": "0:52:30", "remaining_time": "3:55:51"} +{"current_steps": 1370, "total_steps": 7496, "loss": 0.4767, "lr": 0.00019587381318387102, "epoch": 1.4621131270010672, "percentage": 18.28, "elapsed_time": "0:53:05", "remaining_time": "3:57:23"} +{"current_steps": 1375, "total_steps": 7496, "loss": 0.4809, "lr": 0.0001958073566627992, "epoch": 1.4674493062966916, "percentage": 18.34, "elapsed_time": "0:53:37", "remaining_time": "3:58:42"} +{"current_steps": 1380, "total_steps": 7496, "loss": 0.4804, "lr": 0.00019574038068934525, "epoch": 1.472785485592316, "percentage": 18.41, "elapsed_time": "0:54:18", "remaining_time": "4:00:40"} +{"current_steps": 1385, "total_steps": 7496, "loss": 0.4818, "lr": 0.00019567288562664237, "epoch": 1.4781216648879403, "percentage": 18.48, "elapsed_time": "0:54:54", "remaining_time": "4:02:14"} +{"current_steps": 1390, "total_steps": 7496, "loss": 0.4797, "lr": 0.00019560487184063806, "epoch": 1.4834578441835646, "percentage": 18.54, "elapsed_time": "0:55:27", "remaining_time": "4:03:38"} +{"current_steps": 1395, "total_steps": 7496, "loss": 0.4802, "lr": 0.00019553633970009244, "epoch": 1.4887940234791888, "percentage": 18.61, "elapsed_time": "0:56:04", "remaining_time": "4:05:12"} +{"current_steps": 1400, "total_steps": 7496, "loss": 0.4765, "lr": 0.00019546728957657588, "epoch": 1.4941302027748131, "percentage": 18.68, "elapsed_time": "0:56:44", "remaining_time": "4:07:03"} +{"current_steps": 1405, "total_steps": 7496, "loss": 0.4797, "lr": 0.00019539772184446735, "epoch": 1.4994663820704375, "percentage": 18.74, "elapsed_time": "0:57:28", "remaining_time": "4:09:10"} +{"current_steps": 1410, "total_steps": 7496, "loss": 0.4772, "lr": 0.00019532763688095208, "epoch": 1.5048025613660618, "percentage": 18.81, "elapsed_time": "0:58:05", "remaining_time": "4:10:42"} +{"current_steps": 1415, "total_steps": 7496, "loss": 0.4783, "lr": 0.0001952570350660197, "epoch": 1.5101387406616862, "percentage": 18.88, "elapsed_time": "0:58:43", "remaining_time": "4:12:21"} +{"current_steps": 1420, "total_steps": 7496, "loss": 0.4796, "lr": 0.0001951859167824621, "epoch": 1.5154749199573105, "percentage": 18.94, "elapsed_time": "0:59:19", "remaining_time": "4:13:49"} +{"current_steps": 1425, "total_steps": 7496, "loss": 0.48, "lr": 0.00019511428241587143, "epoch": 1.520811099252935, "percentage": 19.01, "elapsed_time": "0:59:53", "remaining_time": "4:15:10"} +{"current_steps": 1430, "total_steps": 7496, "loss": 0.4761, "lr": 0.00019504213235463792, "epoch": 1.5261472785485592, "percentage": 19.08, "elapsed_time": "1:00:31", "remaining_time": "4:16:46"} +{"current_steps": 1435, "total_steps": 7496, "loss": 0.478, "lr": 0.0001949694669899478, "epoch": 1.5314834578441836, "percentage": 19.14, "elapsed_time": "1:01:10", "remaining_time": "4:18:24"} +{"current_steps": 1440, "total_steps": 7496, "loss": 0.478, "lr": 0.00019489628671578126, "epoch": 1.536819637139808, "percentage": 19.21, "elapsed_time": "1:01:46", "remaining_time": "4:19:47"} +{"current_steps": 1445, "total_steps": 7496, "loss": 0.476, "lr": 0.00019482259192891017, "epoch": 1.5421558164354323, "percentage": 19.28, "elapsed_time": "1:02:24", "remaining_time": "4:21:20"} +{"current_steps": 1450, "total_steps": 7496, "loss": 0.4753, "lr": 0.00019474838302889608, "epoch": 1.5474919957310567, "percentage": 19.34, "elapsed_time": "1:03:01", "remaining_time": "4:22:46"} +{"current_steps": 1455, "total_steps": 7496, "loss": 0.4767, "lr": 0.00019467366041808797, "epoch": 1.552828175026681, "percentage": 19.41, "elapsed_time": "1:04:03", "remaining_time": "4:25:56"} +{"current_steps": 1460, "total_steps": 7496, "loss": 0.4783, "lr": 0.00019459842450161998, "epoch": 1.5581643543223054, "percentage": 19.48, "elapsed_time": "1:04:40", "remaining_time": "4:27:24"} +{"current_steps": 1465, "total_steps": 7496, "loss": 0.4758, "lr": 0.00019452267568740946, "epoch": 1.5635005336179295, "percentage": 19.54, "elapsed_time": "1:05:17", "remaining_time": "4:28:46"} +{"current_steps": 1470, "total_steps": 7496, "loss": 0.4767, "lr": 0.00019444641438615446, "epoch": 1.5688367129135539, "percentage": 19.61, "elapsed_time": "1:05:55", "remaining_time": "4:30:14"} +{"current_steps": 1475, "total_steps": 7496, "loss": 0.4748, "lr": 0.00019436964101133178, "epoch": 1.5741728922091782, "percentage": 19.68, "elapsed_time": "1:06:32", "remaining_time": "4:31:37"} +{"current_steps": 1480, "total_steps": 7496, "loss": 0.4743, "lr": 0.00019429235597919457, "epoch": 1.5795090715048026, "percentage": 19.74, "elapsed_time": "1:07:09", "remaining_time": "4:32:58"} +{"current_steps": 1485, "total_steps": 7496, "loss": 0.4748, "lr": 0.00019421455970877006, "epoch": 1.584845250800427, "percentage": 19.81, "elapsed_time": "1:07:47", "remaining_time": "4:34:25"} +{"current_steps": 1490, "total_steps": 7496, "loss": 0.4772, "lr": 0.00019413625262185735, "epoch": 1.590181430096051, "percentage": 19.88, "elapsed_time": "1:08:21", "remaining_time": "4:35:33"} +{"current_steps": 1495, "total_steps": 7496, "loss": 0.4771, "lr": 0.00019405743514302516, "epoch": 1.5955176093916754, "percentage": 19.94, "elapsed_time": "1:09:23", "remaining_time": "4:38:32"} +{"current_steps": 1500, "total_steps": 7496, "loss": 0.4758, "lr": 0.0001939781076996094, "epoch": 1.6008537886872998, "percentage": 20.01, "elapsed_time": "1:10:00", "remaining_time": "4:39:49"} +{"current_steps": 1505, "total_steps": 7496, "loss": 0.4778, "lr": 0.00019389827072171096, "epoch": 1.6061899679829241, "percentage": 20.08, "elapsed_time": "1:10:49", "remaining_time": "4:41:57"} +{"current_steps": 1510, "total_steps": 7496, "loss": 0.477, "lr": 0.0001938179246421934, "epoch": 1.6115261472785485, "percentage": 20.14, "elapsed_time": "1:11:25", "remaining_time": "4:43:09"} +{"current_steps": 1515, "total_steps": 7496, "loss": 0.4749, "lr": 0.0001937370698966804, "epoch": 1.6168623265741728, "percentage": 20.21, "elapsed_time": "1:12:02", "remaining_time": "4:44:25"} +{"current_steps": 1520, "total_steps": 7496, "loss": 0.476, "lr": 0.00019365570692355373, "epoch": 1.6221985058697972, "percentage": 20.28, "elapsed_time": "1:12:38", "remaining_time": "4:45:36"} +{"current_steps": 1525, "total_steps": 7496, "loss": 0.4738, "lr": 0.00019357383616395055, "epoch": 1.6275346851654215, "percentage": 20.34, "elapsed_time": "1:15:03", "remaining_time": "4:53:51"} +{"current_steps": 1530, "total_steps": 7496, "loss": 0.4749, "lr": 0.00019349145806176125, "epoch": 1.632870864461046, "percentage": 20.41, "elapsed_time": "1:15:39", "remaining_time": "4:55:02"} +{"current_steps": 1535, "total_steps": 7496, "loss": 0.4736, "lr": 0.00019340857306362685, "epoch": 1.6382070437566703, "percentage": 20.48, "elapsed_time": "1:16:16", "remaining_time": "4:56:13"} +{"current_steps": 1540, "total_steps": 7496, "loss": 0.4747, "lr": 0.00019332518161893682, "epoch": 1.6435432230522946, "percentage": 20.54, "elapsed_time": "1:18:19", "remaining_time": "5:02:54"} +{"current_steps": 1545, "total_steps": 7496, "loss": 0.4734, "lr": 0.00019324128417982637, "epoch": 1.648879402347919, "percentage": 20.61, "elapsed_time": "1:18:54", "remaining_time": "5:03:57"} +{"current_steps": 1550, "total_steps": 7496, "loss": 0.4729, "lr": 0.0001931568812011742, "epoch": 1.6542155816435433, "percentage": 20.68, "elapsed_time": "1:19:54", "remaining_time": "5:06:32"} +{"current_steps": 1555, "total_steps": 7496, "loss": 0.4741, "lr": 0.00019307197314059996, "epoch": 1.6595517609391677, "percentage": 20.74, "elapsed_time": "1:20:54", "remaining_time": "5:09:07"} +{"current_steps": 1560, "total_steps": 7496, "loss": 0.475, "lr": 0.00019298656045846176, "epoch": 1.664887940234792, "percentage": 20.81, "elapsed_time": "1:22:24", "remaining_time": "5:13:35"} +{"current_steps": 1565, "total_steps": 7496, "loss": 0.4735, "lr": 0.00019290064361785373, "epoch": 1.6702241195304164, "percentage": 20.88, "elapsed_time": "1:23:00", "remaining_time": "5:14:36"} +{"current_steps": 1570, "total_steps": 7496, "loss": 0.4733, "lr": 0.00019281422308460336, "epoch": 1.6755602988260405, "percentage": 20.94, "elapsed_time": "1:23:37", "remaining_time": "5:15:39"} +{"current_steps": 1575, "total_steps": 7496, "loss": 0.4726, "lr": 0.0001927272993272692, "epoch": 1.6808964781216649, "percentage": 21.01, "elapsed_time": "1:24:37", "remaining_time": "5:18:07"} +{"current_steps": 1580, "total_steps": 7496, "loss": 0.4754, "lr": 0.00019263987281713818, "epoch": 1.6862326574172892, "percentage": 21.08, "elapsed_time": "1:25:13", "remaining_time": "5:19:07"} +{"current_steps": 1585, "total_steps": 7496, "loss": 0.4722, "lr": 0.00019255194402822298, "epoch": 1.6915688367129136, "percentage": 21.14, "elapsed_time": "1:26:43", "remaining_time": "5:23:26"} +{"current_steps": 1590, "total_steps": 7496, "loss": 0.4711, "lr": 0.00019246351343725973, "epoch": 1.696905016008538, "percentage": 21.21, "elapsed_time": "1:27:51", "remaining_time": "5:26:22"} +{"current_steps": 1595, "total_steps": 7496, "loss": 0.4706, "lr": 0.00019237458152370507, "epoch": 1.702241195304162, "percentage": 21.28, "elapsed_time": "1:30:12", "remaining_time": "5:33:44"} +{"current_steps": 1600, "total_steps": 7496, "loss": 0.4724, "lr": 0.00019228514876973386, "epoch": 1.7075773745997864, "percentage": 21.34, "elapsed_time": "1:30:48", "remaining_time": "5:34:37"} +{"current_steps": 1605, "total_steps": 7496, "loss": 0.4703, "lr": 0.00019219521566023637, "epoch": 1.7129135538954108, "percentage": 21.41, "elapsed_time": "1:31:38", "remaining_time": "5:36:22"} +{"current_steps": 1610, "total_steps": 7496, "loss": 0.4719, "lr": 0.00019210478268281576, "epoch": 1.7182497331910351, "percentage": 21.48, "elapsed_time": "1:32:15", "remaining_time": "5:37:15"} +{"current_steps": 1615, "total_steps": 7496, "loss": 0.4748, "lr": 0.00019201385032778534, "epoch": 1.7235859124866595, "percentage": 21.54, "elapsed_time": "1:32:51", "remaining_time": "5:38:07"} +{"current_steps": 1620, "total_steps": 7496, "loss": 0.4708, "lr": 0.00019192241908816602, "epoch": 1.7289220917822838, "percentage": 21.61, "elapsed_time": "1:34:16", "remaining_time": "5:41:56"} +{"current_steps": 1625, "total_steps": 7496, "loss": 0.4705, "lr": 0.00019183048945968357, "epoch": 1.7342582710779082, "percentage": 21.68, "elapsed_time": "1:34:55", "remaining_time": "5:42:57"} +{"current_steps": 1630, "total_steps": 7496, "loss": 0.4706, "lr": 0.00019173806194076597, "epoch": 1.7395944503735326, "percentage": 21.74, "elapsed_time": "1:36:15", "remaining_time": "5:46:24"} +{"current_steps": 1635, "total_steps": 7496, "loss": 0.4702, "lr": 0.0001916451370325406, "epoch": 1.744930629669157, "percentage": 21.81, "elapsed_time": "1:37:15", "remaining_time": "5:48:40"} +{"current_steps": 1640, "total_steps": 7496, "loss": 0.4698, "lr": 0.00019155171523883166, "epoch": 1.7502668089647813, "percentage": 21.88, "elapsed_time": "1:38:21", "remaining_time": "5:51:13"} +{"current_steps": 1645, "total_steps": 7496, "loss": 0.4705, "lr": 0.00019145779706615745, "epoch": 1.7556029882604056, "percentage": 21.95, "elapsed_time": "1:39:20", "remaining_time": "5:53:19"} +{"current_steps": 1650, "total_steps": 7496, "loss": 0.4687, "lr": 0.00019136338302372746, "epoch": 1.76093916755603, "percentage": 22.01, "elapsed_time": "1:40:20", "remaining_time": "5:55:32"} +{"current_steps": 1655, "total_steps": 7496, "loss": 0.4684, "lr": 0.0001912684736234397, "epoch": 1.7662753468516543, "percentage": 22.08, "elapsed_time": "1:41:22", "remaining_time": "5:57:46"} +{"current_steps": 1660, "total_steps": 7496, "loss": 0.4705, "lr": 0.00019117306937987803, "epoch": 1.7716115261472787, "percentage": 22.15, "elapsed_time": "1:41:59", "remaining_time": "5:58:34"} +{"current_steps": 1665, "total_steps": 7496, "loss": 0.4681, "lr": 0.00019107717081030918, "epoch": 1.776947705442903, "percentage": 22.21, "elapsed_time": "1:42:36", "remaining_time": "5:59:19"} +{"current_steps": 1670, "total_steps": 7496, "loss": 0.4685, "lr": 0.00019098077843468012, "epoch": 1.7822838847385272, "percentage": 22.28, "elapsed_time": "1:43:10", "remaining_time": "5:59:55"} +{"current_steps": 1675, "total_steps": 7496, "loss": 0.469, "lr": 0.00019088389277561508, "epoch": 1.7876200640341515, "percentage": 22.35, "elapsed_time": "1:43:45", "remaining_time": "6:00:35"} +{"current_steps": 1680, "total_steps": 7496, "loss": 0.4689, "lr": 0.00019078651435841285, "epoch": 1.7929562433297759, "percentage": 22.41, "elapsed_time": "1:44:50", "remaining_time": "6:02:55"} +{"current_steps": 1685, "total_steps": 7496, "loss": 0.4671, "lr": 0.0001906886437110438, "epoch": 1.7982924226254002, "percentage": 22.48, "elapsed_time": "1:45:25", "remaining_time": "6:03:35"} +{"current_steps": 1690, "total_steps": 7496, "loss": 0.4707, "lr": 0.0001905902813641472, "epoch": 1.8036286019210246, "percentage": 22.55, "elapsed_time": "1:46:03", "remaining_time": "6:04:21"} +{"current_steps": 1695, "total_steps": 7496, "loss": 0.4692, "lr": 0.00019049142785102817, "epoch": 1.8089647812166487, "percentage": 22.61, "elapsed_time": "1:46:41", "remaining_time": "6:05:07"} +{"current_steps": 1700, "total_steps": 7496, "loss": 0.4676, "lr": 0.00019039208370765488, "epoch": 1.814300960512273, "percentage": 22.68, "elapsed_time": "1:47:19", "remaining_time": "6:05:55"} +{"current_steps": 1705, "total_steps": 7496, "loss": 0.4676, "lr": 0.00019029224947265561, "epoch": 1.8196371398078974, "percentage": 22.75, "elapsed_time": "1:48:05", "remaining_time": "6:07:08"} +{"current_steps": 1710, "total_steps": 7496, "loss": 0.4673, "lr": 0.00019019192568731582, "epoch": 1.8249733191035218, "percentage": 22.81, "elapsed_time": "1:48:41", "remaining_time": "6:07:45"} +{"current_steps": 1715, "total_steps": 7496, "loss": 0.4652, "lr": 0.0001900911128955753, "epoch": 1.8303094983991461, "percentage": 22.88, "elapsed_time": "1:49:17", "remaining_time": "6:08:25"} +{"current_steps": 1720, "total_steps": 7496, "loss": 0.4654, "lr": 0.0001899898116440251, "epoch": 1.8356456776947705, "percentage": 22.95, "elapsed_time": "1:49:54", "remaining_time": "6:09:03"} +{"current_steps": 1725, "total_steps": 7496, "loss": 0.4675, "lr": 0.00018988802248190458, "epoch": 1.8409818569903948, "percentage": 23.01, "elapsed_time": "1:50:51", "remaining_time": "6:10:53"} +{"current_steps": 1730, "total_steps": 7496, "loss": 0.4668, "lr": 0.0001897857459610986, "epoch": 1.8463180362860192, "percentage": 23.08, "elapsed_time": "1:51:28", "remaining_time": "6:11:32"} +{"current_steps": 1735, "total_steps": 7496, "loss": 0.4687, "lr": 0.0001896829826361343, "epoch": 1.8516542155816436, "percentage": 23.15, "elapsed_time": "1:52:05", "remaining_time": "6:12:10"} +{"current_steps": 1740, "total_steps": 7496, "loss": 0.4672, "lr": 0.0001895797330641782, "epoch": 1.856990394877268, "percentage": 23.21, "elapsed_time": "1:52:41", "remaining_time": "6:12:46"} +{"current_steps": 1745, "total_steps": 7496, "loss": 0.4661, "lr": 0.00018947599780503324, "epoch": 1.8623265741728923, "percentage": 23.28, "elapsed_time": "1:53:16", "remaining_time": "6:13:20"} +{"current_steps": 1750, "total_steps": 7496, "loss": 0.4665, "lr": 0.0001893717774211356, "epoch": 1.8676627534685166, "percentage": 23.35, "elapsed_time": "1:53:54", "remaining_time": "6:13:59"} +{"current_steps": 1755, "total_steps": 7496, "loss": 0.4658, "lr": 0.0001892670724775518, "epoch": 1.872998932764141, "percentage": 23.41, "elapsed_time": "1:54:30", "remaining_time": "6:14:34"} +{"current_steps": 1760, "total_steps": 7496, "loss": 0.4654, "lr": 0.00018916188354197559, "epoch": 1.8783351120597653, "percentage": 23.48, "elapsed_time": "1:55:06", "remaining_time": "6:15:08"} +{"current_steps": 1765, "total_steps": 7496, "loss": 0.4669, "lr": 0.00018905621118472468, "epoch": 1.8836712913553897, "percentage": 23.55, "elapsed_time": "1:55:43", "remaining_time": "6:15:44"} +{"current_steps": 1770, "total_steps": 7496, "loss": 0.4658, "lr": 0.000188950055978738, "epoch": 1.8890074706510138, "percentage": 23.61, "elapsed_time": "1:56:19", "remaining_time": "6:16:18"} +{"current_steps": 1775, "total_steps": 7496, "loss": 0.4678, "lr": 0.0001888434184995723, "epoch": 1.8943436499466382, "percentage": 23.68, "elapsed_time": "1:57:16", "remaining_time": "6:17:58"} +{"current_steps": 1780, "total_steps": 7496, "loss": 0.4646, "lr": 0.00018873629932539921, "epoch": 1.8996798292422625, "percentage": 23.75, "elapsed_time": "1:58:15", "remaining_time": "6:19:46"} +{"current_steps": 1785, "total_steps": 7496, "loss": 0.4636, "lr": 0.000188628699037002, "epoch": 1.9050160085378869, "percentage": 23.81, "elapsed_time": "1:59:20", "remaining_time": "6:21:50"} +{"current_steps": 1790, "total_steps": 7496, "loss": 0.4647, "lr": 0.00018852061821777248, "epoch": 1.9103521878335112, "percentage": 23.88, "elapsed_time": "1:59:55", "remaining_time": "6:22:18"} +{"current_steps": 1795, "total_steps": 7496, "loss": 0.4639, "lr": 0.00018841205745370775, "epoch": 1.9156883671291356, "percentage": 23.95, "elapsed_time": "2:00:32", "remaining_time": "6:22:49"} +{"current_steps": 1800, "total_steps": 7496, "loss": 0.4657, "lr": 0.00018830301733340724, "epoch": 1.9210245464247597, "percentage": 24.01, "elapsed_time": "2:01:35", "remaining_time": "6:24:46"} +{"current_steps": 1805, "total_steps": 7496, "loss": 0.4659, "lr": 0.00018819349844806924, "epoch": 1.926360725720384, "percentage": 24.08, "elapsed_time": "2:02:23", "remaining_time": "6:25:54"} +{"current_steps": 1810, "total_steps": 7496, "loss": 0.4619, "lr": 0.0001880835013914879, "epoch": 1.9316969050160084, "percentage": 24.15, "elapsed_time": "2:03:00", "remaining_time": "6:26:24"} +{"current_steps": 1815, "total_steps": 7496, "loss": 0.4647, "lr": 0.00018797302676004988, "epoch": 1.9370330843116328, "percentage": 24.21, "elapsed_time": "2:03:38", "remaining_time": "6:27:00"} +{"current_steps": 1820, "total_steps": 7496, "loss": 0.4668, "lr": 0.00018786207515273127, "epoch": 1.9423692636072571, "percentage": 24.28, "elapsed_time": "2:04:16", "remaining_time": "6:27:35"} +{"current_steps": 1825, "total_steps": 7496, "loss": 0.4669, "lr": 0.00018775064717109415, "epoch": 1.9477054429028815, "percentage": 24.35, "elapsed_time": "2:04:50", "remaining_time": "6:27:56"} +{"current_steps": 1830, "total_steps": 7496, "loss": 0.4645, "lr": 0.00018763874341928344, "epoch": 1.9530416221985059, "percentage": 24.41, "elapsed_time": "2:05:27", "remaining_time": "6:28:26"} +{"current_steps": 1835, "total_steps": 7496, "loss": 0.4651, "lr": 0.00018752636450402373, "epoch": 1.9583778014941302, "percentage": 24.48, "elapsed_time": "2:06:06", "remaining_time": "6:29:04"} +{"current_steps": 1840, "total_steps": 7496, "loss": 0.4637, "lr": 0.0001874135110346157, "epoch": 1.9637139807897546, "percentage": 24.55, "elapsed_time": "2:06:42", "remaining_time": "6:29:29"} +{"current_steps": 1845, "total_steps": 7496, "loss": 0.4651, "lr": 0.0001873001836229331, "epoch": 1.969050160085379, "percentage": 24.61, "elapsed_time": "2:07:17", "remaining_time": "6:29:53"} +{"current_steps": 1850, "total_steps": 7496, "loss": 0.4639, "lr": 0.00018718638288341933, "epoch": 1.9743863393810033, "percentage": 24.68, "elapsed_time": "2:07:55", "remaining_time": "6:30:25"} +{"current_steps": 1855, "total_steps": 7496, "loss": 0.4655, "lr": 0.00018707210943308397, "epoch": 1.9797225186766276, "percentage": 24.75, "elapsed_time": "2:08:30", "remaining_time": "6:30:48"} +{"current_steps": 1860, "total_steps": 7496, "loss": 0.4623, "lr": 0.0001869573638914997, "epoch": 1.985058697972252, "percentage": 24.81, "elapsed_time": "2:09:06", "remaining_time": "6:31:12"} +{"current_steps": 1865, "total_steps": 7496, "loss": 0.4646, "lr": 0.00018684214688079875, "epoch": 1.9903948772678763, "percentage": 24.88, "elapsed_time": "2:09:43", "remaining_time": "6:31:41"} +{"current_steps": 1870, "total_steps": 7496, "loss": 0.4623, "lr": 0.00018672645902566957, "epoch": 1.9957310565635007, "percentage": 24.95, "elapsed_time": "2:10:19", "remaining_time": "6:32:05"} +{"current_steps": 1875, "total_steps": 7496, "loss": 0.4622, "lr": 0.0001866103009533535, "epoch": 2.001067235859125, "percentage": 25.01, "elapsed_time": "2:10:54", "remaining_time": "6:32:25"} +{"current_steps": 1880, "total_steps": 7496, "loss": 0.4608, "lr": 0.00018649367329364128, "epoch": 2.0064034151547494, "percentage": 25.08, "elapsed_time": "2:11:37", "remaining_time": "6:33:11"} +{"current_steps": 1885, "total_steps": 7496, "loss": 0.4607, "lr": 0.00018637657667886965, "epoch": 2.0117395944503733, "percentage": 25.15, "elapsed_time": "2:12:20", "remaining_time": "6:33:55"} +{"current_steps": 1890, "total_steps": 7496, "loss": 0.4599, "lr": 0.0001862590117439181, "epoch": 2.0170757737459977, "percentage": 25.21, "elapsed_time": "2:13:00", "remaining_time": "6:34:32"} +{"current_steps": 1895, "total_steps": 7496, "loss": 0.4607, "lr": 0.00018614097912620506, "epoch": 2.022411953041622, "percentage": 25.28, "elapsed_time": "2:13:36", "remaining_time": "6:34:52"} +{"current_steps": 1900, "total_steps": 7496, "loss": 0.4587, "lr": 0.0001860224794656848, "epoch": 2.0277481323372464, "percentage": 25.35, "elapsed_time": "2:14:14", "remaining_time": "6:35:22"} +{"current_steps": 1905, "total_steps": 7496, "loss": 0.4595, "lr": 0.00018590351340484388, "epoch": 2.0330843116328707, "percentage": 25.41, "elapsed_time": "2:15:03", "remaining_time": "6:36:22"} +{"current_steps": 1910, "total_steps": 7496, "loss": 0.4583, "lr": 0.0001857840815886974, "epoch": 2.038420490928495, "percentage": 25.48, "elapsed_time": "2:15:37", "remaining_time": "6:36:38"} +{"current_steps": 1915, "total_steps": 7496, "loss": 0.4594, "lr": 0.0001856641846647859, "epoch": 2.0437566702241194, "percentage": 25.55, "elapsed_time": "2:16:15", "remaining_time": "6:37:07"} +{"current_steps": 1920, "total_steps": 7496, "loss": 0.4599, "lr": 0.00018554382328317163, "epoch": 2.049092849519744, "percentage": 25.61, "elapsed_time": "2:16:54", "remaining_time": "6:37:37"} +{"current_steps": 1925, "total_steps": 7496, "loss": 0.4601, "lr": 0.00018542299809643493, "epoch": 2.054429028815368, "percentage": 25.68, "elapsed_time": "2:17:28", "remaining_time": "6:37:52"} +{"current_steps": 1930, "total_steps": 7496, "loss": 0.4604, "lr": 0.00018530170975967098, "epoch": 2.0597652081109925, "percentage": 25.75, "elapsed_time": "2:18:06", "remaining_time": "6:38:16"} +{"current_steps": 1935, "total_steps": 7496, "loss": 0.4604, "lr": 0.00018517995893048604, "epoch": 2.065101387406617, "percentage": 25.81, "elapsed_time": "2:18:46", "remaining_time": "6:38:48"} +{"current_steps": 1940, "total_steps": 7496, "loss": 0.4588, "lr": 0.0001850577462689939, "epoch": 2.070437566702241, "percentage": 25.88, "elapsed_time": "2:19:20", "remaining_time": "6:39:04"} +{"current_steps": 1945, "total_steps": 7496, "loss": 0.4612, "lr": 0.00018493507243781224, "epoch": 2.0757737459978656, "percentage": 25.95, "elapsed_time": "2:19:56", "remaining_time": "6:39:23"} +{"current_steps": 1950, "total_steps": 7496, "loss": 0.4586, "lr": 0.00018481193810205933, "epoch": 2.08110992529349, "percentage": 26.01, "elapsed_time": "2:20:36", "remaining_time": "6:39:53"} +{"current_steps": 1955, "total_steps": 7496, "loss": 0.4576, "lr": 0.00018468834392935016, "epoch": 2.0864461045891143, "percentage": 26.08, "elapsed_time": "2:21:11", "remaining_time": "6:40:09"} +{"current_steps": 1960, "total_steps": 7496, "loss": 0.4601, "lr": 0.00018456429058979278, "epoch": 2.0917822838847386, "percentage": 26.15, "elapsed_time": "2:21:45", "remaining_time": "6:40:24"} +{"current_steps": 1965, "total_steps": 7496, "loss": 0.4574, "lr": 0.0001844397787559848, "epoch": 2.097118463180363, "percentage": 26.21, "elapsed_time": "2:22:26", "remaining_time": "6:40:56"} +{"current_steps": 1970, "total_steps": 7496, "loss": 0.456, "lr": 0.00018431480910300984, "epoch": 2.1024546424759873, "percentage": 26.28, "elapsed_time": "2:23:02", "remaining_time": "6:41:14"} +{"current_steps": 1975, "total_steps": 7496, "loss": 0.4578, "lr": 0.00018418938230843358, "epoch": 2.1077908217716117, "percentage": 26.35, "elapsed_time": "2:23:37", "remaining_time": "6:41:29"} +{"current_steps": 1980, "total_steps": 7496, "loss": 0.4581, "lr": 0.00018406349905230037, "epoch": 2.113127001067236, "percentage": 26.41, "elapsed_time": "2:24:16", "remaining_time": "6:41:56"} +{"current_steps": 1985, "total_steps": 7496, "loss": 0.4587, "lr": 0.0001839371600171294, "epoch": 2.11846318036286, "percentage": 26.48, "elapsed_time": "2:24:52", "remaining_time": "6:42:13"} +{"current_steps": 1990, "total_steps": 7496, "loss": 0.4591, "lr": 0.00018381036588791094, "epoch": 2.1237993596584843, "percentage": 26.55, "elapsed_time": "2:25:26", "remaining_time": "6:42:24"} +{"current_steps": 1995, "total_steps": 7496, "loss": 0.4587, "lr": 0.0001836831173521028, "epoch": 2.1291355389541087, "percentage": 26.61, "elapsed_time": "2:26:03", "remaining_time": "6:42:43"} +{"current_steps": 2000, "total_steps": 7496, "loss": 0.4567, "lr": 0.00018355541509962645, "epoch": 2.134471718249733, "percentage": 26.68, "elapsed_time": "2:26:41", "remaining_time": "6:43:05"} +{"current_steps": 2005, "total_steps": 7496, "loss": 0.4576, "lr": 0.00018342725982286342, "epoch": 2.1398078975453574, "percentage": 26.75, "elapsed_time": "2:27:26", "remaining_time": "6:43:47"} +{"current_steps": 2010, "total_steps": 7496, "loss": 0.4579, "lr": 0.00018329865221665138, "epoch": 2.1451440768409817, "percentage": 26.81, "elapsed_time": "2:28:02", "remaining_time": "6:44:04"} +{"current_steps": 2015, "total_steps": 7496, "loss": 0.4571, "lr": 0.0001831695929782805, "epoch": 2.150480256136606, "percentage": 26.88, "elapsed_time": "2:28:41", "remaining_time": "6:44:28"} +{"current_steps": 2020, "total_steps": 7496, "loss": 0.4577, "lr": 0.00018304008280748963, "epoch": 2.1558164354322304, "percentage": 26.95, "elapsed_time": "2:29:16", "remaining_time": "6:44:39"} +{"current_steps": 2025, "total_steps": 7496, "loss": 0.4588, "lr": 0.00018291012240646246, "epoch": 2.161152614727855, "percentage": 27.01, "elapsed_time": "2:29:51", "remaining_time": "6:44:53"} +{"current_steps": 2030, "total_steps": 7496, "loss": 0.4576, "lr": 0.00018277971247982383, "epoch": 2.166488794023479, "percentage": 27.08, "elapsed_time": "2:30:32", "remaining_time": "6:45:20"} +{"current_steps": 2035, "total_steps": 7496, "loss": 0.4549, "lr": 0.0001826488537346358, "epoch": 2.1718249733191035, "percentage": 27.15, "elapsed_time": "2:31:08", "remaining_time": "6:45:34"} +{"current_steps": 2040, "total_steps": 7496, "loss": 0.4565, "lr": 0.00018251754688039385, "epoch": 2.177161152614728, "percentage": 27.21, "elapsed_time": "2:31:42", "remaining_time": "6:45:45"} +{"current_steps": 2045, "total_steps": 7496, "loss": 0.4566, "lr": 0.00018238579262902301, "epoch": 2.1824973319103522, "percentage": 27.28, "elapsed_time": "2:32:19", "remaining_time": "6:46:02"} +{"current_steps": 2050, "total_steps": 7496, "loss": 0.4568, "lr": 0.00018225359169487412, "epoch": 2.1878335112059766, "percentage": 27.35, "elapsed_time": "2:32:57", "remaining_time": "6:46:21"} +{"current_steps": 2055, "total_steps": 7496, "loss": 0.457, "lr": 0.00018212094479471975, "epoch": 2.193169690501601, "percentage": 27.41, "elapsed_time": "2:33:32", "remaining_time": "6:46:31"} +{"current_steps": 2060, "total_steps": 7496, "loss": 0.4567, "lr": 0.0001819878526477505, "epoch": 2.1985058697972253, "percentage": 27.48, "elapsed_time": "2:34:06", "remaining_time": "6:46:41"} +{"current_steps": 2065, "total_steps": 7496, "loss": 0.4561, "lr": 0.000181854315975571, "epoch": 2.2038420490928496, "percentage": 27.55, "elapsed_time": "2:34:46", "remaining_time": "6:47:04"} +{"current_steps": 2070, "total_steps": 7496, "loss": 0.4552, "lr": 0.000181720335502196, "epoch": 2.209178228388474, "percentage": 27.61, "elapsed_time": "2:35:24", "remaining_time": "6:47:20"} +{"current_steps": 2075, "total_steps": 7496, "loss": 0.4561, "lr": 0.0001815859119540466, "epoch": 2.2145144076840984, "percentage": 27.68, "elapsed_time": "2:35:59", "remaining_time": "6:47:30"} +{"current_steps": 2080, "total_steps": 7496, "loss": 0.459, "lr": 0.00018145104605994599, "epoch": 2.2198505869797227, "percentage": 27.75, "elapsed_time": "2:36:37", "remaining_time": "6:47:49"} +{"current_steps": 2085, "total_steps": 7496, "loss": 0.4563, "lr": 0.00018131573855111578, "epoch": 2.225186766275347, "percentage": 27.81, "elapsed_time": "2:37:39", "remaining_time": "6:49:08"} +{"current_steps": 2090, "total_steps": 7496, "loss": 0.4556, "lr": 0.00018117999016117203, "epoch": 2.2305229455709714, "percentage": 27.88, "elapsed_time": "2:38:15", "remaining_time": "6:49:20"} +{"current_steps": 2095, "total_steps": 7496, "loss": 0.4565, "lr": 0.00018104380162612101, "epoch": 2.2358591248665953, "percentage": 27.95, "elapsed_time": "2:38:53", "remaining_time": "6:49:38"} +{"current_steps": 2100, "total_steps": 7496, "loss": 0.4555, "lr": 0.0001809071736843556, "epoch": 2.2411953041622197, "percentage": 28.01, "elapsed_time": "2:39:29", "remaining_time": "6:49:48"} +{"current_steps": 2105, "total_steps": 7496, "loss": 0.4549, "lr": 0.00018077010707665084, "epoch": 2.246531483457844, "percentage": 28.08, "elapsed_time": "2:40:17", "remaining_time": "6:50:29"} +{"current_steps": 2110, "total_steps": 7496, "loss": 0.4563, "lr": 0.00018063260254616033, "epoch": 2.2518676627534684, "percentage": 28.15, "elapsed_time": "2:40:55", "remaining_time": "6:50:45"} +{"current_steps": 2115, "total_steps": 7496, "loss": 0.4558, "lr": 0.00018049466083841194, "epoch": 2.2572038420490927, "percentage": 28.22, "elapsed_time": "2:41:31", "remaining_time": "6:50:57"} +{"current_steps": 2120, "total_steps": 7496, "loss": 0.4546, "lr": 0.0001803562827013039, "epoch": 2.262540021344717, "percentage": 28.28, "elapsed_time": "2:42:07", "remaining_time": "6:51:07"} +{"current_steps": 2125, "total_steps": 7496, "loss": 0.4576, "lr": 0.00018021746888510057, "epoch": 2.2678762006403415, "percentage": 28.35, "elapsed_time": "2:42:43", "remaining_time": "6:51:16"} +{"current_steps": 2130, "total_steps": 7496, "loss": 0.4553, "lr": 0.00018007822014242867, "epoch": 2.273212379935966, "percentage": 28.42, "elapsed_time": "2:43:19", "remaining_time": "6:51:28"} +{"current_steps": 2135, "total_steps": 7496, "loss": 0.4544, "lr": 0.00017993853722827284, "epoch": 2.27854855923159, "percentage": 28.48, "elapsed_time": "2:44:43", "remaining_time": "6:53:38"} +{"current_steps": 2140, "total_steps": 7496, "loss": 0.4561, "lr": 0.00017979842089997186, "epoch": 2.2838847385272145, "percentage": 28.55, "elapsed_time": "2:45:19", "remaining_time": "6:53:47"} +{"current_steps": 2145, "total_steps": 7496, "loss": 0.4558, "lr": 0.00017965787191721435, "epoch": 2.289220917822839, "percentage": 28.62, "elapsed_time": "2:45:57", "remaining_time": "6:53:59"} +{"current_steps": 2150, "total_steps": 7496, "loss": 0.4527, "lr": 0.0001795168910420348, "epoch": 2.2945570971184632, "percentage": 28.68, "elapsed_time": "2:46:37", "remaining_time": "6:54:18"} +{"current_steps": 2155, "total_steps": 7496, "loss": 0.4544, "lr": 0.00017937547903880917, "epoch": 2.2998932764140876, "percentage": 28.75, "elapsed_time": "2:47:14", "remaining_time": "6:54:28"} +{"current_steps": 2160, "total_steps": 7496, "loss": 0.4529, "lr": 0.00017923363667425116, "epoch": 2.305229455709712, "percentage": 28.82, "elapsed_time": "2:47:51", "remaining_time": "6:54:41"} +{"current_steps": 2165, "total_steps": 7496, "loss": 0.4552, "lr": 0.00017909136471740765, "epoch": 2.3105656350053363, "percentage": 28.88, "elapsed_time": "2:48:28", "remaining_time": "6:54:51"} +{"current_steps": 2170, "total_steps": 7496, "loss": 0.4539, "lr": 0.00017894866393965476, "epoch": 2.3159018143009606, "percentage": 28.95, "elapsed_time": "2:49:05", "remaining_time": "6:55:01"} +{"current_steps": 2175, "total_steps": 7496, "loss": 0.4543, "lr": 0.0001788055351146936, "epoch": 2.321237993596585, "percentage": 29.02, "elapsed_time": "2:49:44", "remaining_time": "6:55:15"} +{"current_steps": 2180, "total_steps": 7496, "loss": 0.455, "lr": 0.00017866197901854615, "epoch": 2.3265741728922094, "percentage": 29.08, "elapsed_time": "2:50:20", "remaining_time": "6:55:21"} +{"current_steps": 2185, "total_steps": 7496, "loss": 0.4568, "lr": 0.00017851799642955088, "epoch": 2.3319103521878333, "percentage": 29.15, "elapsed_time": "2:50:57", "remaining_time": "6:55:33"} +{"current_steps": 2190, "total_steps": 7496, "loss": 0.457, "lr": 0.00017837358812835866, "epoch": 2.3372465314834576, "percentage": 29.22, "elapsed_time": "2:51:37", "remaining_time": "6:55:48"} +{"current_steps": 2195, "total_steps": 7496, "loss": 0.4542, "lr": 0.0001782287548979286, "epoch": 2.342582710779082, "percentage": 29.28, "elapsed_time": "2:52:13", "remaining_time": "6:55:56"} +{"current_steps": 2200, "total_steps": 7496, "loss": 0.4551, "lr": 0.00017808349752352357, "epoch": 2.3479188900747063, "percentage": 29.35, "elapsed_time": "2:52:50", "remaining_time": "6:56:03"} +{"current_steps": 2205, "total_steps": 7496, "loss": 0.4553, "lr": 0.0001779378167927062, "epoch": 2.3532550693703307, "percentage": 29.42, "elapsed_time": "2:53:38", "remaining_time": "6:56:40"} +{"current_steps": 2210, "total_steps": 7496, "loss": 0.4542, "lr": 0.00017779171349533446, "epoch": 2.358591248665955, "percentage": 29.48, "elapsed_time": "2:54:17", "remaining_time": "6:56:52"} +{"current_steps": 2215, "total_steps": 7496, "loss": 0.4542, "lr": 0.0001776451884235573, "epoch": 2.3639274279615794, "percentage": 29.55, "elapsed_time": "2:54:53", "remaining_time": "6:56:58"} +{"current_steps": 2220, "total_steps": 7496, "loss": 0.4528, "lr": 0.00017749824237181066, "epoch": 2.3692636072572038, "percentage": 29.62, "elapsed_time": "2:55:30", "remaining_time": "6:57:05"} +{"current_steps": 2225, "total_steps": 7496, "loss": 0.4535, "lr": 0.00017735087613681284, "epoch": 2.374599786552828, "percentage": 29.68, "elapsed_time": "2:56:07", "remaining_time": "6:57:14"} +{"current_steps": 2230, "total_steps": 7496, "loss": 0.4537, "lr": 0.00017720309051756026, "epoch": 2.3799359658484525, "percentage": 29.75, "elapsed_time": "2:57:11", "remaining_time": "6:58:26"} +{"current_steps": 2235, "total_steps": 7496, "loss": 0.4532, "lr": 0.00017705488631532333, "epoch": 2.385272145144077, "percentage": 29.82, "elapsed_time": "2:57:50", "remaining_time": "6:58:37"} +{"current_steps": 2240, "total_steps": 7496, "loss": 0.4504, "lr": 0.00017690626433364184, "epoch": 2.390608324439701, "percentage": 29.88, "elapsed_time": "2:58:54", "remaining_time": "6:59:47"} +{"current_steps": 2245, "total_steps": 7496, "loss": 0.4545, "lr": 0.00017675722537832073, "epoch": 2.3959445037353255, "percentage": 29.95, "elapsed_time": "2:59:33", "remaining_time": "6:59:59"} +{"current_steps": 2250, "total_steps": 7496, "loss": 0.4526, "lr": 0.00017660777025742572, "epoch": 2.40128068303095, "percentage": 30.02, "elapsed_time": "3:00:09", "remaining_time": "7:00:03"} +{"current_steps": 2255, "total_steps": 7496, "loss": 0.4519, "lr": 0.00017645789978127893, "epoch": 2.4066168623265742, "percentage": 30.08, "elapsed_time": "3:01:18", "remaining_time": "7:01:23"} +{"current_steps": 2260, "total_steps": 7496, "loss": 0.4541, "lr": 0.00017630761476245446, "epoch": 2.4119530416221986, "percentage": 30.15, "elapsed_time": "3:01:55", "remaining_time": "7:01:29"} +{"current_steps": 2265, "total_steps": 7496, "loss": 0.4538, "lr": 0.00017615691601577393, "epoch": 2.417289220917823, "percentage": 30.22, "elapsed_time": "3:02:34", "remaining_time": "7:01:39"} +{"current_steps": 2270, "total_steps": 7496, "loss": 0.4539, "lr": 0.00017600580435830226, "epoch": 2.4226254002134473, "percentage": 30.28, "elapsed_time": "3:03:11", "remaining_time": "7:01:44"} +{"current_steps": 2275, "total_steps": 7496, "loss": 0.4513, "lr": 0.00017585428060934295, "epoch": 2.4279615795090717, "percentage": 30.35, "elapsed_time": "3:03:46", "remaining_time": "7:01:46"} +{"current_steps": 2280, "total_steps": 7496, "loss": 0.4533, "lr": 0.00017570234559043392, "epoch": 2.433297758804696, "percentage": 30.42, "elapsed_time": "3:04:24", "remaining_time": "7:01:51"} +{"current_steps": 2285, "total_steps": 7496, "loss": 0.4509, "lr": 0.00017555000012534292, "epoch": 2.4386339381003204, "percentage": 30.48, "elapsed_time": "3:05:00", "remaining_time": "7:01:55"} +{"current_steps": 2290, "total_steps": 7496, "loss": 0.4542, "lr": 0.000175397245040063, "epoch": 2.4439701173959447, "percentage": 30.55, "elapsed_time": "3:05:35", "remaining_time": "7:01:54"} +{"current_steps": 2295, "total_steps": 7496, "loss": 0.4533, "lr": 0.00017524408116280813, "epoch": 2.449306296691569, "percentage": 30.62, "elapsed_time": "3:06:11", "remaining_time": "7:01:57"} +{"current_steps": 2300, "total_steps": 7496, "loss": 0.4529, "lr": 0.00017509050932400876, "epoch": 2.454642475987193, "percentage": 30.68, "elapsed_time": "3:06:50", "remaining_time": "7:02:05"} +{"current_steps": 2305, "total_steps": 7496, "loss": 0.4529, "lr": 0.00017493653035630715, "epoch": 2.4599786552828173, "percentage": 30.75, "elapsed_time": "3:07:41", "remaining_time": "7:02:40"} +{"current_steps": 2310, "total_steps": 7496, "loss": 0.4506, "lr": 0.000174782145094553, "epoch": 2.4653148345784417, "percentage": 30.82, "elapsed_time": "3:08:42", "remaining_time": "7:03:38"} +{"current_steps": 2315, "total_steps": 7496, "loss": 0.4544, "lr": 0.00017462735437579884, "epoch": 2.470651013874066, "percentage": 30.88, "elapsed_time": "3:09:18", "remaining_time": "7:03:39"} +{"current_steps": 2320, "total_steps": 7496, "loss": 0.452, "lr": 0.00017447215903929552, "epoch": 2.4759871931696904, "percentage": 30.95, "elapsed_time": "3:10:18", "remaining_time": "7:04:34"} +{"current_steps": 2325, "total_steps": 7496, "loss": 0.452, "lr": 0.00017431655992648776, "epoch": 2.4813233724653148, "percentage": 31.02, "elapsed_time": "3:10:54", "remaining_time": "7:04:35"} +{"current_steps": 2330, "total_steps": 7496, "loss": 0.4517, "lr": 0.00017416055788100935, "epoch": 2.486659551760939, "percentage": 31.08, "elapsed_time": "3:12:52", "remaining_time": "7:07:37"} +{"current_steps": 2335, "total_steps": 7496, "loss": 0.4517, "lr": 0.0001740041537486788, "epoch": 2.4919957310565635, "percentage": 31.15, "elapsed_time": "3:13:29", "remaining_time": "7:07:39"} +{"current_steps": 2340, "total_steps": 7496, "loss": 0.4505, "lr": 0.00017384734837749472, "epoch": 2.497331910352188, "percentage": 31.22, "elapsed_time": "3:14:03", "remaining_time": "7:07:35"} +{"current_steps": 2345, "total_steps": 7496, "loss": 0.4509, "lr": 0.000173690142617631, "epoch": 2.502668089647812, "percentage": 31.28, "elapsed_time": "3:14:41", "remaining_time": "7:07:40"} +{"current_steps": 2350, "total_steps": 7496, "loss": 0.4518, "lr": 0.00017353253732143255, "epoch": 2.5080042689434365, "percentage": 31.35, "elapsed_time": "3:15:20", "remaining_time": "7:07:46"} +{"current_steps": 2355, "total_steps": 7496, "loss": 0.4512, "lr": 0.00017337453334341044, "epoch": 2.513340448239061, "percentage": 31.42, "elapsed_time": "3:15:56", "remaining_time": "7:07:43"} +{"current_steps": 2360, "total_steps": 7496, "loss": 0.4513, "lr": 0.00017321613154023727, "epoch": 2.5186766275346852, "percentage": 31.48, "elapsed_time": "3:17:00", "remaining_time": "7:08:43"} +{"current_steps": 2365, "total_steps": 7496, "loss": 0.4495, "lr": 0.00017305733277074272, "epoch": 2.5240128068303096, "percentage": 31.55, "elapsed_time": "3:18:29", "remaining_time": "7:10:37"} +{"current_steps": 2370, "total_steps": 7496, "loss": 0.4522, "lr": 0.00017289813789590864, "epoch": 2.529348986125934, "percentage": 31.62, "elapsed_time": "3:19:40", "remaining_time": "7:11:52"} +{"current_steps": 2375, "total_steps": 7496, "loss": 0.4505, "lr": 0.0001727385477788645, "epoch": 2.5346851654215583, "percentage": 31.68, "elapsed_time": "3:20:17", "remaining_time": "7:11:51"} +{"current_steps": 2380, "total_steps": 7496, "loss": 0.4531, "lr": 0.00017257856328488285, "epoch": 2.5400213447171827, "percentage": 31.75, "elapsed_time": "3:20:57", "remaining_time": "7:11:58"} +{"current_steps": 2385, "total_steps": 7496, "loss": 0.4503, "lr": 0.0001724181852813743, "epoch": 2.5453575240128066, "percentage": 31.82, "elapsed_time": "3:21:34", "remaining_time": "7:11:57"} +{"current_steps": 2390, "total_steps": 7496, "loss": 0.4505, "lr": 0.00017225741463788315, "epoch": 2.550693703308431, "percentage": 31.88, "elapsed_time": "3:22:09", "remaining_time": "7:11:54"} +{"current_steps": 2395, "total_steps": 7496, "loss": 0.4521, "lr": 0.0001720962522260825, "epoch": 2.5560298826040553, "percentage": 31.95, "elapsed_time": "3:22:51", "remaining_time": "7:12:04"} +{"current_steps": 2400, "total_steps": 7496, "loss": 0.4494, "lr": 0.00017193469891976947, "epoch": 2.5613660618996796, "percentage": 32.02, "elapsed_time": "3:23:28", "remaining_time": "7:12:01"} +{"current_steps": 2405, "total_steps": 7496, "loss": 0.4521, "lr": 0.00017177275559486068, "epoch": 2.566702241195304, "percentage": 32.08, "elapsed_time": "3:24:12", "remaining_time": "7:12:15"} +{"current_steps": 2410, "total_steps": 7496, "loss": 0.4509, "lr": 0.00017161042312938723, "epoch": 2.5720384204909283, "percentage": 32.15, "elapsed_time": "3:24:51", "remaining_time": "7:12:20"} +{"current_steps": 2415, "total_steps": 7496, "loss": 0.45, "lr": 0.00017144770240349022, "epoch": 2.5773745997865527, "percentage": 32.22, "elapsed_time": "3:25:55", "remaining_time": "7:13:14"} +{"current_steps": 2420, "total_steps": 7496, "loss": 0.4485, "lr": 0.00017128459429941576, "epoch": 2.582710779082177, "percentage": 32.28, "elapsed_time": "3:26:31", "remaining_time": "7:13:11"} +{"current_steps": 2425, "total_steps": 7496, "loss": 0.4516, "lr": 0.00017112109970151025, "epoch": 2.5880469583778014, "percentage": 32.35, "elapsed_time": "3:27:09", "remaining_time": "7:13:10"} +{"current_steps": 2430, "total_steps": 7496, "loss": 0.4486, "lr": 0.00017095721949621565, "epoch": 2.5933831376734258, "percentage": 32.42, "elapsed_time": "3:27:47", "remaining_time": "7:13:12"} +{"current_steps": 2435, "total_steps": 7496, "loss": 0.4504, "lr": 0.00017079295457206456, "epoch": 2.59871931696905, "percentage": 32.48, "elapsed_time": "3:28:25", "remaining_time": "7:13:12"} +{"current_steps": 2440, "total_steps": 7496, "loss": 0.448, "lr": 0.0001706283058196755, "epoch": 2.6040554962646745, "percentage": 32.55, "elapsed_time": "3:29:01", "remaining_time": "7:13:06"} +{"current_steps": 2445, "total_steps": 7496, "loss": 0.4482, "lr": 0.0001704632741317481, "epoch": 2.609391675560299, "percentage": 32.62, "elapsed_time": "3:29:37", "remaining_time": "7:13:03"} +{"current_steps": 2450, "total_steps": 7496, "loss": 0.4498, "lr": 0.00017029786040305807, "epoch": 2.614727854855923, "percentage": 32.68, "elapsed_time": "3:30:44", "remaining_time": "7:14:01"} +{"current_steps": 2455, "total_steps": 7496, "loss": 0.4499, "lr": 0.00017013206553045268, "epoch": 2.6200640341515475, "percentage": 32.75, "elapsed_time": "3:31:20", "remaining_time": "7:13:56"} +{"current_steps": 2460, "total_steps": 7496, "loss": 0.4485, "lr": 0.00016996589041284546, "epoch": 2.625400213447172, "percentage": 32.82, "elapsed_time": "3:31:56", "remaining_time": "7:13:52"} +{"current_steps": 2465, "total_steps": 7496, "loss": 0.4473, "lr": 0.00016979933595121177, "epoch": 2.6307363927427962, "percentage": 32.88, "elapsed_time": "3:32:36", "remaining_time": "7:13:55"} +{"current_steps": 2470, "total_steps": 7496, "loss": 0.4469, "lr": 0.00016963240304858362, "epoch": 2.6360725720384206, "percentage": 32.95, "elapsed_time": "3:33:15", "remaining_time": "7:13:57"} +{"current_steps": 2475, "total_steps": 7496, "loss": 0.4497, "lr": 0.00016946509261004495, "epoch": 2.641408751334045, "percentage": 33.02, "elapsed_time": "3:34:16", "remaining_time": "7:14:41"} +{"current_steps": 2480, "total_steps": 7496, "loss": 0.4473, "lr": 0.00016929740554272646, "epoch": 2.6467449306296693, "percentage": 33.08, "elapsed_time": "3:35:45", "remaining_time": "7:16:22"} +{"current_steps": 2485, "total_steps": 7496, "loss": 0.4499, "lr": 0.0001691293427558011, "epoch": 2.6520811099252937, "percentage": 33.15, "elapsed_time": "3:36:24", "remaining_time": "7:16:22"} +{"current_steps": 2490, "total_steps": 7496, "loss": 0.4494, "lr": 0.00016896090516047872, "epoch": 2.657417289220918, "percentage": 33.22, "elapsed_time": "3:37:24", "remaining_time": "7:17:05"} +{"current_steps": 2495, "total_steps": 7496, "loss": 0.4482, "lr": 0.0001687920936700015, "epoch": 2.6627534685165424, "percentage": 33.28, "elapsed_time": "3:38:37", "remaining_time": "7:18:12"} +{"current_steps": 2500, "total_steps": 7496, "loss": 0.4482, "lr": 0.00016862290919963862, "epoch": 2.6680896478121667, "percentage": 33.35, "elapsed_time": "3:39:13", "remaining_time": "7:18:06"} +{"current_steps": 2505, "total_steps": 7496, "loss": 0.447, "lr": 0.00016845335266668166, "epoch": 2.673425827107791, "percentage": 33.42, "elapsed_time": "3:40:28", "remaining_time": "7:19:16"} +{"current_steps": 2510, "total_steps": 7496, "loss": 0.4482, "lr": 0.0001682834249904394, "epoch": 2.678762006403415, "percentage": 33.48, "elapsed_time": "3:42:00", "remaining_time": "7:21:01"} +{"current_steps": 2515, "total_steps": 7496, "loss": 0.4484, "lr": 0.00016811312709223293, "epoch": 2.6840981856990394, "percentage": 33.55, "elapsed_time": "3:42:36", "remaining_time": "7:20:53"} +{"current_steps": 2520, "total_steps": 7496, "loss": 0.448, "lr": 0.00016794245989539054, "epoch": 2.6894343649946637, "percentage": 33.62, "elapsed_time": "3:43:13", "remaining_time": "7:20:47"} +{"current_steps": 2525, "total_steps": 7496, "loss": 0.4457, "lr": 0.00016777142432524293, "epoch": 2.694770544290288, "percentage": 33.68, "elapsed_time": "3:43:51", "remaining_time": "7:20:43"} +{"current_steps": 2530, "total_steps": 7496, "loss": 0.4472, "lr": 0.00016760002130911797, "epoch": 2.7001067235859124, "percentage": 33.75, "elapsed_time": "3:44:29", "remaining_time": "7:20:39"} +{"current_steps": 2535, "total_steps": 7496, "loss": 0.4464, "lr": 0.0001674282517763358, "epoch": 2.7054429028815368, "percentage": 33.82, "elapsed_time": "3:45:08", "remaining_time": "7:20:35"} +{"current_steps": 2540, "total_steps": 7496, "loss": 0.4464, "lr": 0.0001672561166582037, "epoch": 2.710779082177161, "percentage": 33.88, "elapsed_time": "3:45:45", "remaining_time": "7:20:29"} +{"current_steps": 2545, "total_steps": 7496, "loss": 0.4474, "lr": 0.0001670836168880112, "epoch": 2.7161152614727855, "percentage": 33.95, "elapsed_time": "3:46:21", "remaining_time": "7:20:20"} +{"current_steps": 2550, "total_steps": 7496, "loss": 0.4451, "lr": 0.00016691075340102485, "epoch": 2.72145144076841, "percentage": 34.02, "elapsed_time": "3:47:00", "remaining_time": "7:20:19"} +{"current_steps": 2555, "total_steps": 7496, "loss": 0.448, "lr": 0.00016673752713448314, "epoch": 2.726787620064034, "percentage": 34.08, "elapsed_time": "3:47:39", "remaining_time": "7:20:15"} +{"current_steps": 2560, "total_steps": 7496, "loss": 0.4475, "lr": 0.00016656393902759166, "epoch": 2.7321237993596585, "percentage": 34.15, "elapsed_time": "3:48:14", "remaining_time": "7:20:04"} +{"current_steps": 2565, "total_steps": 7496, "loss": 0.447, "lr": 0.00016638999002151775, "epoch": 2.737459978655283, "percentage": 34.22, "elapsed_time": "3:48:52", "remaining_time": "7:19:59"} +{"current_steps": 2570, "total_steps": 7496, "loss": 0.4464, "lr": 0.00016621568105938548, "epoch": 2.7427961579509073, "percentage": 34.28, "elapsed_time": "3:49:29", "remaining_time": "7:19:53"} +{"current_steps": 2575, "total_steps": 7496, "loss": 0.4465, "lr": 0.00016604101308627053, "epoch": 2.7481323372465316, "percentage": 34.35, "elapsed_time": "3:50:04", "remaining_time": "7:19:41"} +{"current_steps": 2580, "total_steps": 7496, "loss": 0.4476, "lr": 0.00016586598704919516, "epoch": 2.753468516542156, "percentage": 34.42, "elapsed_time": "3:50:41", "remaining_time": "7:19:34"} +{"current_steps": 2585, "total_steps": 7496, "loss": 0.4462, "lr": 0.0001656906038971229, "epoch": 2.75880469583778, "percentage": 34.49, "elapsed_time": "3:51:20", "remaining_time": "7:19:29"} +{"current_steps": 2590, "total_steps": 7496, "loss": 0.4447, "lr": 0.0001655148645809536, "epoch": 2.7641408751334042, "percentage": 34.55, "elapsed_time": "3:51:57", "remaining_time": "7:19:22"} +{"current_steps": 2595, "total_steps": 7496, "loss": 0.4458, "lr": 0.0001653387700535181, "epoch": 2.7694770544290286, "percentage": 34.62, "elapsed_time": "3:52:35", "remaining_time": "7:19:15"} +{"current_steps": 2600, "total_steps": 7496, "loss": 0.445, "lr": 0.00016516232126957312, "epoch": 2.774813233724653, "percentage": 34.69, "elapsed_time": "3:53:37", "remaining_time": "7:19:55"} +{"current_steps": 2605, "total_steps": 7496, "loss": 0.4464, "lr": 0.00016498551918579622, "epoch": 2.7801494130202773, "percentage": 34.75, "elapsed_time": "3:55:09", "remaining_time": "7:21:31"} +{"current_steps": 2610, "total_steps": 7496, "loss": 0.4466, "lr": 0.00016480836476078038, "epoch": 2.7854855923159016, "percentage": 34.82, "elapsed_time": "3:55:45", "remaining_time": "7:21:21"} +{"current_steps": 2615, "total_steps": 7496, "loss": 0.449, "lr": 0.00016463085895502893, "epoch": 2.790821771611526, "percentage": 34.89, "elapsed_time": "3:56:47", "remaining_time": "7:21:59"} +{"current_steps": 2620, "total_steps": 7496, "loss": 0.448, "lr": 0.00016445300273095048, "epoch": 2.7961579509071504, "percentage": 34.95, "elapsed_time": "3:57:24", "remaining_time": "7:21:49"} +{"current_steps": 2625, "total_steps": 7496, "loss": 0.4457, "lr": 0.0001642747970528533, "epoch": 2.8014941302027747, "percentage": 35.02, "elapsed_time": "3:57:59", "remaining_time": "7:21:38"} +{"current_steps": 2630, "total_steps": 7496, "loss": 0.4472, "lr": 0.00016409624288694056, "epoch": 2.806830309498399, "percentage": 35.09, "elapsed_time": "3:58:38", "remaining_time": "7:21:31"} +{"current_steps": 2635, "total_steps": 7496, "loss": 0.4463, "lr": 0.00016391734120130478, "epoch": 2.8121664887940234, "percentage": 35.15, "elapsed_time": "3:59:15", "remaining_time": "7:21:22"} +{"current_steps": 2640, "total_steps": 7496, "loss": 0.4447, "lr": 0.00016373809296592266, "epoch": 2.8175026680896478, "percentage": 35.22, "elapsed_time": "3:59:50", "remaining_time": "7:21:10"} +{"current_steps": 2645, "total_steps": 7496, "loss": 0.449, "lr": 0.00016355849915264989, "epoch": 2.822838847385272, "percentage": 35.29, "elapsed_time": "4:00:26", "remaining_time": "7:20:58"} +{"current_steps": 2650, "total_steps": 7496, "loss": 0.4467, "lr": 0.00016337856073521585, "epoch": 2.8281750266808965, "percentage": 35.35, "elapsed_time": "4:01:04", "remaining_time": "7:20:51"} +{"current_steps": 2655, "total_steps": 7496, "loss": 0.4449, "lr": 0.00016319827868921822, "epoch": 2.833511205976521, "percentage": 35.42, "elapsed_time": "4:01:41", "remaining_time": "7:20:41"} +{"current_steps": 2660, "total_steps": 7496, "loss": 0.4467, "lr": 0.0001630176539921178, "epoch": 2.838847385272145, "percentage": 35.49, "elapsed_time": "4:02:18", "remaining_time": "7:20:32"} +{"current_steps": 2665, "total_steps": 7496, "loss": 0.4439, "lr": 0.0001628366876232333, "epoch": 2.8441835645677696, "percentage": 35.55, "elapsed_time": "4:03:22", "remaining_time": "7:21:10"} +{"current_steps": 2670, "total_steps": 7496, "loss": 0.4465, "lr": 0.00016265538056373582, "epoch": 2.849519743863394, "percentage": 35.62, "elapsed_time": "4:04:48", "remaining_time": "7:22:29"} +{"current_steps": 2675, "total_steps": 7496, "loss": 0.4457, "lr": 0.00016247373379664362, "epoch": 2.8548559231590183, "percentage": 35.69, "elapsed_time": "4:05:25", "remaining_time": "7:22:19"} +{"current_steps": 2680, "total_steps": 7496, "loss": 0.4445, "lr": 0.00016229174830681683, "epoch": 2.8601921024546426, "percentage": 35.75, "elapsed_time": "4:06:30", "remaining_time": "7:22:57"} +{"current_steps": 2685, "total_steps": 7496, "loss": 0.4426, "lr": 0.00016210942508095212, "epoch": 2.865528281750267, "percentage": 35.82, "elapsed_time": "4:07:08", "remaining_time": "7:22:49"} +{"current_steps": 2690, "total_steps": 7496, "loss": 0.4457, "lr": 0.00016192676510757718, "epoch": 2.8708644610458913, "percentage": 35.89, "elapsed_time": "4:07:47", "remaining_time": "7:22:43"} +{"current_steps": 2695, "total_steps": 7496, "loss": 0.4438, "lr": 0.0001617437693770457, "epoch": 2.8762006403415157, "percentage": 35.95, "elapsed_time": "4:08:49", "remaining_time": "7:23:15"} +{"current_steps": 2700, "total_steps": 7496, "loss": 0.4441, "lr": 0.00016156043888153164, "epoch": 2.88153681963714, "percentage": 36.02, "elapsed_time": "4:09:26", "remaining_time": "7:23:05"} +{"current_steps": 2705, "total_steps": 7496, "loss": 0.4446, "lr": 0.00016137677461502403, "epoch": 2.8868729989327644, "percentage": 36.09, "elapsed_time": "4:11:07", "remaining_time": "7:24:46"} +{"current_steps": 2710, "total_steps": 7496, "loss": 0.443, "lr": 0.00016119277757332157, "epoch": 2.8922091782283887, "percentage": 36.15, "elapsed_time": "4:11:44", "remaining_time": "7:24:35"} +{"current_steps": 2715, "total_steps": 7496, "loss": 0.4449, "lr": 0.0001610084487540272, "epoch": 2.8975453575240127, "percentage": 36.22, "elapsed_time": "4:12:49", "remaining_time": "7:25:12"} +{"current_steps": 2720, "total_steps": 7496, "loss": 0.4464, "lr": 0.00016082378915654275, "epoch": 2.902881536819637, "percentage": 36.29, "elapsed_time": "4:13:26", "remaining_time": "7:25:00"} +{"current_steps": 2725, "total_steps": 7496, "loss": 0.4458, "lr": 0.0001606387997820634, "epoch": 2.9082177161152614, "percentage": 36.35, "elapsed_time": "4:14:03", "remaining_time": "7:24:48"} +{"current_steps": 2730, "total_steps": 7496, "loss": 0.4464, "lr": 0.0001604534816335724, "epoch": 2.9135538954108857, "percentage": 36.42, "elapsed_time": "4:14:39", "remaining_time": "7:24:34"} +{"current_steps": 2735, "total_steps": 7496, "loss": 0.4441, "lr": 0.0001602678357158355, "epoch": 2.91889007470651, "percentage": 36.49, "elapsed_time": "4:15:17", "remaining_time": "7:24:23"} +{"current_steps": 2740, "total_steps": 7496, "loss": 0.4438, "lr": 0.00016008186303539566, "epoch": 2.9242262540021344, "percentage": 36.55, "elapsed_time": "4:15:54", "remaining_time": "7:24:12"} +{"current_steps": 2745, "total_steps": 7496, "loss": 0.4432, "lr": 0.00015989556460056735, "epoch": 2.929562433297759, "percentage": 36.62, "elapsed_time": "4:16:30", "remaining_time": "7:23:58"} +{"current_steps": 2750, "total_steps": 7496, "loss": 0.4423, "lr": 0.00015970894142143137, "epoch": 2.934898612593383, "percentage": 36.69, "elapsed_time": "4:17:08", "remaining_time": "7:23:46"} +{"current_steps": 2755, "total_steps": 7496, "loss": 0.4408, "lr": 0.00015952199450982915, "epoch": 2.9402347918890075, "percentage": 36.75, "elapsed_time": "4:17:48", "remaining_time": "7:23:39"} +{"current_steps": 2760, "total_steps": 7496, "loss": 0.4448, "lr": 0.0001593347248793574, "epoch": 2.945570971184632, "percentage": 36.82, "elapsed_time": "4:18:25", "remaining_time": "7:23:26"} +{"current_steps": 2765, "total_steps": 7496, "loss": 0.4448, "lr": 0.00015914713354536243, "epoch": 2.950907150480256, "percentage": 36.89, "elapsed_time": "4:19:03", "remaining_time": "7:23:15"} +{"current_steps": 2770, "total_steps": 7496, "loss": 0.4438, "lr": 0.000158959221524935, "epoch": 2.9562433297758806, "percentage": 36.95, "elapsed_time": "4:20:04", "remaining_time": "7:23:43"} +{"current_steps": 2775, "total_steps": 7496, "loss": 0.4432, "lr": 0.00015877098983690442, "epoch": 2.961579509071505, "percentage": 37.02, "elapsed_time": "4:21:13", "remaining_time": "7:24:24"} +{"current_steps": 2780, "total_steps": 7496, "loss": 0.4438, "lr": 0.00015858243950183317, "epoch": 2.9669156883671293, "percentage": 37.09, "elapsed_time": "4:21:51", "remaining_time": "7:24:12"} +{"current_steps": 2785, "total_steps": 7496, "loss": 0.4437, "lr": 0.0001583935715420115, "epoch": 2.9722518676627536, "percentage": 37.15, "elapsed_time": "4:22:28", "remaining_time": "7:23:59"} +{"current_steps": 2790, "total_steps": 7496, "loss": 0.4424, "lr": 0.0001582043869814517, "epoch": 2.9775880469583775, "percentage": 37.22, "elapsed_time": "4:23:07", "remaining_time": "7:23:48"} +{"current_steps": 2795, "total_steps": 7496, "loss": 0.4418, "lr": 0.0001580148868458826, "epoch": 2.982924226254002, "percentage": 37.29, "elapsed_time": "4:23:43", "remaining_time": "7:23:33"} +{"current_steps": 2800, "total_steps": 7496, "loss": 0.4434, "lr": 0.0001578250721627441, "epoch": 2.9882604055496262, "percentage": 37.35, "elapsed_time": "4:24:18", "remaining_time": "7:23:16"} +{"current_steps": 2805, "total_steps": 7496, "loss": 0.4423, "lr": 0.00015763494396118144, "epoch": 2.9935965848452506, "percentage": 37.42, "elapsed_time": "4:25:05", "remaining_time": "7:23:20"} +{"current_steps": 2810, "total_steps": 7496, "loss": 0.4427, "lr": 0.00015744450327203976, "epoch": 2.998932764140875, "percentage": 37.49, "elapsed_time": "4:25:41", "remaining_time": "7:23:03"} +{"current_steps": 2815, "total_steps": 7496, "loss": 0.4358, "lr": 0.00015725375112785845, "epoch": 3.0042689434364993, "percentage": 37.55, "elapsed_time": "4:26:18", "remaining_time": "7:22:49"} +{"current_steps": 2820, "total_steps": 7496, "loss": 0.4375, "lr": 0.0001570626885628656, "epoch": 3.0096051227321237, "percentage": 37.62, "elapsed_time": "4:26:56", "remaining_time": "7:22:38"} +{"current_steps": 2825, "total_steps": 7496, "loss": 0.439, "lr": 0.00015687131661297227, "epoch": 3.014941302027748, "percentage": 37.69, "elapsed_time": "4:27:35", "remaining_time": "7:22:27"} +{"current_steps": 2830, "total_steps": 7496, "loss": 0.4377, "lr": 0.00015667963631576704, "epoch": 3.0202774813233724, "percentage": 37.75, "elapsed_time": "4:28:14", "remaining_time": "7:22:16"} +{"current_steps": 2835, "total_steps": 7496, "loss": 0.4372, "lr": 0.00015648764871051022, "epoch": 3.0256136606189967, "percentage": 37.82, "elapsed_time": "4:28:52", "remaining_time": "7:22:03"} +{"current_steps": 2840, "total_steps": 7496, "loss": 0.4374, "lr": 0.00015629535483812834, "epoch": 3.030949839914621, "percentage": 37.89, "elapsed_time": "4:29:29", "remaining_time": "7:21:48"} +{"current_steps": 2845, "total_steps": 7496, "loss": 0.437, "lr": 0.00015610275574120847, "epoch": 3.0362860192102454, "percentage": 37.95, "elapsed_time": "4:30:08", "remaining_time": "7:21:37"} +{"current_steps": 2850, "total_steps": 7496, "loss": 0.4385, "lr": 0.00015590985246399254, "epoch": 3.04162219850587, "percentage": 38.02, "elapsed_time": "4:30:44", "remaining_time": "7:21:21"} +{"current_steps": 2855, "total_steps": 7496, "loss": 0.4392, "lr": 0.0001557166460523717, "epoch": 3.046958377801494, "percentage": 38.09, "elapsed_time": "4:31:21", "remaining_time": "7:21:07"} +{"current_steps": 2860, "total_steps": 7496, "loss": 0.4407, "lr": 0.00015552313755388072, "epoch": 3.0522945570971185, "percentage": 38.15, "elapsed_time": "4:32:00", "remaining_time": "7:20:55"} +{"current_steps": 2865, "total_steps": 7496, "loss": 0.4384, "lr": 0.00015532932801769203, "epoch": 3.057630736392743, "percentage": 38.22, "elapsed_time": "4:32:39", "remaining_time": "7:20:43"} +{"current_steps": 2870, "total_steps": 7496, "loss": 0.4374, "lr": 0.0001551352184946105, "epoch": 3.062966915688367, "percentage": 38.29, "elapsed_time": "4:33:44", "remaining_time": "7:21:14"} +{"current_steps": 2875, "total_steps": 7496, "loss": 0.439, "lr": 0.00015494081003706732, "epoch": 3.0683030949839916, "percentage": 38.35, "elapsed_time": "4:35:04", "remaining_time": "7:22:08"} +{"current_steps": 2880, "total_steps": 7496, "loss": 0.4359, "lr": 0.00015474610369911447, "epoch": 3.073639274279616, "percentage": 38.42, "elapsed_time": "4:36:16", "remaining_time": "7:22:48"} +{"current_steps": 2885, "total_steps": 7496, "loss": 0.4365, "lr": 0.00015455110053641904, "epoch": 3.0789754535752403, "percentage": 38.49, "elapsed_time": "4:36:54", "remaining_time": "7:22:34"} +{"current_steps": 2890, "total_steps": 7496, "loss": 0.4361, "lr": 0.00015435580160625746, "epoch": 3.0843116328708646, "percentage": 38.55, "elapsed_time": "4:37:31", "remaining_time": "7:22:18"} +{"current_steps": 2895, "total_steps": 7496, "loss": 0.4375, "lr": 0.00015416020796750964, "epoch": 3.089647812166489, "percentage": 38.62, "elapsed_time": "4:38:31", "remaining_time": "7:22:39"} +{"current_steps": 2900, "total_steps": 7496, "loss": 0.4347, "lr": 0.00015396432068065356, "epoch": 3.0949839914621133, "percentage": 38.69, "elapsed_time": "4:39:10", "remaining_time": "7:22:27"} +{"current_steps": 2905, "total_steps": 7496, "loss": 0.4391, "lr": 0.00015376814080775915, "epoch": 3.1003201707577372, "percentage": 38.75, "elapsed_time": "4:39:59", "remaining_time": "7:22:29"} +{"current_steps": 2910, "total_steps": 7496, "loss": 0.4376, "lr": 0.00015357166941248277, "epoch": 3.1056563500533616, "percentage": 38.82, "elapsed_time": "4:41:01", "remaining_time": "7:22:52"} +{"current_steps": 2915, "total_steps": 7496, "loss": 0.4381, "lr": 0.00015337490756006132, "epoch": 3.110992529348986, "percentage": 38.89, "elapsed_time": "4:41:39", "remaining_time": "7:22:37"} +{"current_steps": 2920, "total_steps": 7496, "loss": 0.4389, "lr": 0.00015317785631730661, "epoch": 3.1163287086446103, "percentage": 38.95, "elapsed_time": "4:43:35", "remaining_time": "7:24:24"} +{"current_steps": 2925, "total_steps": 7496, "loss": 0.4376, "lr": 0.00015298051675259933, "epoch": 3.1216648879402347, "percentage": 39.02, "elapsed_time": "4:44:13", "remaining_time": "7:24:10"} +{"current_steps": 2930, "total_steps": 7496, "loss": 0.4347, "lr": 0.00015278288993588346, "epoch": 3.127001067235859, "percentage": 39.09, "elapsed_time": "4:44:51", "remaining_time": "7:23:54"} +{"current_steps": 2935, "total_steps": 7496, "loss": 0.4351, "lr": 0.00015258497693866057, "epoch": 3.1323372465314834, "percentage": 39.15, "elapsed_time": "4:45:29", "remaining_time": "7:23:39"} +{"current_steps": 2940, "total_steps": 7496, "loss": 0.4376, "lr": 0.00015238677883398362, "epoch": 3.1376734258271077, "percentage": 39.22, "elapsed_time": "4:46:06", "remaining_time": "7:23:21"} +{"current_steps": 2945, "total_steps": 7496, "loss": 0.4392, "lr": 0.0001521882966964515, "epoch": 3.143009605122732, "percentage": 39.29, "elapsed_time": "4:46:43", "remaining_time": "7:23:05"} +{"current_steps": 2950, "total_steps": 7496, "loss": 0.4354, "lr": 0.00015198953160220312, "epoch": 3.1483457844183564, "percentage": 39.35, "elapsed_time": "4:47:48", "remaining_time": "7:23:30"} +{"current_steps": 2955, "total_steps": 7496, "loss": 0.4385, "lr": 0.00015179048462891145, "epoch": 3.153681963713981, "percentage": 39.42, "elapsed_time": "4:48:25", "remaining_time": "7:23:13"} +{"current_steps": 2960, "total_steps": 7496, "loss": 0.4353, "lr": 0.00015159115685577774, "epoch": 3.159018143009605, "percentage": 39.49, "elapsed_time": "4:49:01", "remaining_time": "7:22:54"} +{"current_steps": 2965, "total_steps": 7496, "loss": 0.4346, "lr": 0.00015139154936352585, "epoch": 3.1643543223052295, "percentage": 39.55, "elapsed_time": "4:49:39", "remaining_time": "7:22:38"} +{"current_steps": 2970, "total_steps": 7496, "loss": 0.4351, "lr": 0.00015119166323439604, "epoch": 3.169690501600854, "percentage": 39.62, "elapsed_time": "4:51:09", "remaining_time": "7:23:41"} +{"current_steps": 2975, "total_steps": 7496, "loss": 0.438, "lr": 0.00015099149955213943, "epoch": 3.175026680896478, "percentage": 39.69, "elapsed_time": "4:51:46", "remaining_time": "7:23:23"} +{"current_steps": 2980, "total_steps": 7496, "loss": 0.4362, "lr": 0.00015079105940201198, "epoch": 3.1803628601921026, "percentage": 39.75, "elapsed_time": "4:52:48", "remaining_time": "7:23:43"} +{"current_steps": 2985, "total_steps": 7496, "loss": 0.4393, "lr": 0.00015059034387076854, "epoch": 3.185699039487727, "percentage": 39.82, "elapsed_time": "4:54:45", "remaining_time": "7:25:26"} +{"current_steps": 2990, "total_steps": 7496, "loss": 0.4357, "lr": 0.00015038935404665705, "epoch": 3.1910352187833513, "percentage": 39.89, "elapsed_time": "4:55:23", "remaining_time": "7:25:09"} +{"current_steps": 2995, "total_steps": 7496, "loss": 0.4376, "lr": 0.00015018809101941268, "epoch": 3.1963713980789756, "percentage": 39.95, "elapsed_time": "4:56:53", "remaining_time": "7:26:10"} +{"current_steps": 3000, "total_steps": 7496, "loss": 0.4366, "lr": 0.0001499865558802518, "epoch": 3.2017075773746, "percentage": 40.02, "elapsed_time": "4:57:57", "remaining_time": "7:26:32"} +{"current_steps": 3005, "total_steps": 7496, "loss": 0.436, "lr": 0.00014978474972186621, "epoch": 3.207043756670224, "percentage": 40.09, "elapsed_time": "4:59:11", "remaining_time": "7:27:09"} +{"current_steps": 3010, "total_steps": 7496, "loss": 0.4357, "lr": 0.00014958267363841707, "epoch": 3.2123799359658483, "percentage": 40.15, "elapsed_time": "5:00:16", "remaining_time": "7:27:31"} +{"current_steps": 3015, "total_steps": 7496, "loss": 0.4409, "lr": 0.00014938032872552904, "epoch": 3.2177161152614726, "percentage": 40.22, "elapsed_time": "5:00:54", "remaining_time": "7:27:12"} +{"current_steps": 3020, "total_steps": 7496, "loss": 0.4347, "lr": 0.0001491777160802843, "epoch": 3.223052294557097, "percentage": 40.29, "elapsed_time": "5:01:30", "remaining_time": "7:26:51"} +{"current_steps": 3025, "total_steps": 7496, "loss": 0.435, "lr": 0.00014897483680121669, "epoch": 3.2283884738527213, "percentage": 40.35, "elapsed_time": "5:02:05", "remaining_time": "7:26:29"} +{"current_steps": 3030, "total_steps": 7496, "loss": 0.4368, "lr": 0.0001487716919883057, "epoch": 3.2337246531483457, "percentage": 40.42, "elapsed_time": "5:02:41", "remaining_time": "7:26:09"} +{"current_steps": 3035, "total_steps": 7496, "loss": 0.4377, "lr": 0.00014856828274297044, "epoch": 3.23906083244397, "percentage": 40.49, "elapsed_time": "5:03:22", "remaining_time": "7:25:54"} +{"current_steps": 3040, "total_steps": 7496, "loss": 0.4378, "lr": 0.00014836461016806378, "epoch": 3.2443970117395944, "percentage": 40.55, "elapsed_time": "5:03:57", "remaining_time": "7:25:31"} +{"current_steps": 3045, "total_steps": 7496, "loss": 0.436, "lr": 0.0001481606753678663, "epoch": 3.2497331910352187, "percentage": 40.62, "elapsed_time": "5:04:34", "remaining_time": "7:25:12"} +{"current_steps": 3050, "total_steps": 7496, "loss": 0.4364, "lr": 0.00014795647944808036, "epoch": 3.255069370330843, "percentage": 40.69, "elapsed_time": "5:05:13", "remaining_time": "7:24:55"} +{"current_steps": 3055, "total_steps": 7496, "loss": 0.4351, "lr": 0.000147752023515824, "epoch": 3.2604055496264674, "percentage": 40.76, "elapsed_time": "5:05:52", "remaining_time": "7:24:38"} +{"current_steps": 3060, "total_steps": 7496, "loss": 0.4365, "lr": 0.00014754730867962506, "epoch": 3.265741728922092, "percentage": 40.82, "elapsed_time": "5:06:55", "remaining_time": "7:24:56"} +{"current_steps": 3065, "total_steps": 7496, "loss": 0.4363, "lr": 0.00014734233604941517, "epoch": 3.271077908217716, "percentage": 40.89, "elapsed_time": "5:07:35", "remaining_time": "7:24:40"} +{"current_steps": 3070, "total_steps": 7496, "loss": 0.4328, "lr": 0.00014713710673652355, "epoch": 3.2764140875133405, "percentage": 40.96, "elapsed_time": "5:08:38", "remaining_time": "7:24:57"} +{"current_steps": 3075, "total_steps": 7496, "loss": 0.4368, "lr": 0.00014693162185367114, "epoch": 3.281750266808965, "percentage": 41.02, "elapsed_time": "5:09:14", "remaining_time": "7:24:35"} +{"current_steps": 3080, "total_steps": 7496, "loss": 0.4355, "lr": 0.00014672588251496467, "epoch": 3.287086446104589, "percentage": 41.09, "elapsed_time": "5:09:53", "remaining_time": "7:24:19"} +{"current_steps": 3085, "total_steps": 7496, "loss": 0.433, "lr": 0.00014651988983589034, "epoch": 3.2924226254002136, "percentage": 41.16, "elapsed_time": "5:10:30", "remaining_time": "7:23:58"} +{"current_steps": 3090, "total_steps": 7496, "loss": 0.4348, "lr": 0.00014631364493330798, "epoch": 3.297758804695838, "percentage": 41.22, "elapsed_time": "5:11:05", "remaining_time": "7:23:34"} +{"current_steps": 3095, "total_steps": 7496, "loss": 0.4348, "lr": 0.00014610714892544492, "epoch": 3.3030949839914623, "percentage": 41.29, "elapsed_time": "5:11:43", "remaining_time": "7:23:15"} +{"current_steps": 3100, "total_steps": 7496, "loss": 0.4364, "lr": 0.00014590040293189, "epoch": 3.3084311632870866, "percentage": 41.36, "elapsed_time": "5:12:19", "remaining_time": "7:22:54"} +{"current_steps": 3105, "total_steps": 7496, "loss": 0.4347, "lr": 0.0001456934080735874, "epoch": 3.313767342582711, "percentage": 41.42, "elapsed_time": "5:13:05", "remaining_time": "7:22:46"} +{"current_steps": 3110, "total_steps": 7496, "loss": 0.436, "lr": 0.00014548616547283057, "epoch": 3.3191035218783353, "percentage": 41.49, "elapsed_time": "5:13:40", "remaining_time": "7:22:22"} +{"current_steps": 3115, "total_steps": 7496, "loss": 0.4347, "lr": 0.0001452786762532562, "epoch": 3.3244397011739593, "percentage": 41.56, "elapsed_time": "5:14:18", "remaining_time": "7:22:02"} +{"current_steps": 3120, "total_steps": 7496, "loss": 0.4352, "lr": 0.00014507094153983817, "epoch": 3.3297758804695836, "percentage": 41.62, "elapsed_time": "5:15:47", "remaining_time": "7:22:55"} +{"current_steps": 3125, "total_steps": 7496, "loss": 0.434, "lr": 0.00014486296245888133, "epoch": 3.335112059765208, "percentage": 41.69, "elapsed_time": "5:16:22", "remaining_time": "7:22:31"} +{"current_steps": 3130, "total_steps": 7496, "loss": 0.4329, "lr": 0.00014465474013801547, "epoch": 3.3404482390608323, "percentage": 41.76, "elapsed_time": "5:17:55", "remaining_time": "7:23:28"} +{"current_steps": 3135, "total_steps": 7496, "loss": 0.4341, "lr": 0.00014444627570618917, "epoch": 3.3457844183564567, "percentage": 41.82, "elapsed_time": "5:18:31", "remaining_time": "7:23:05"} +{"current_steps": 3140, "total_steps": 7496, "loss": 0.4338, "lr": 0.00014423757029366371, "epoch": 3.351120597652081, "percentage": 41.89, "elapsed_time": "5:19:07", "remaining_time": "7:22:42"} +{"current_steps": 3145, "total_steps": 7496, "loss": 0.4331, "lr": 0.00014402862503200693, "epoch": 3.3564567769477054, "percentage": 41.96, "elapsed_time": "5:19:45", "remaining_time": "7:22:22"} +{"current_steps": 3150, "total_steps": 7496, "loss": 0.4331, "lr": 0.00014381944105408713, "epoch": 3.3617929562433297, "percentage": 42.02, "elapsed_time": "5:20:23", "remaining_time": "7:22:02"} +{"current_steps": 3155, "total_steps": 7496, "loss": 0.4345, "lr": 0.00014361001949406683, "epoch": 3.367129135538954, "percentage": 42.09, "elapsed_time": "5:20:59", "remaining_time": "7:21:39"} +{"current_steps": 3160, "total_steps": 7496, "loss": 0.4366, "lr": 0.00014340036148739667, "epoch": 3.3724653148345785, "percentage": 42.16, "elapsed_time": "5:21:39", "remaining_time": "7:21:21"} +{"current_steps": 3165, "total_steps": 7496, "loss": 0.4326, "lr": 0.0001431904681708094, "epoch": 3.377801494130203, "percentage": 42.22, "elapsed_time": "5:22:16", "remaining_time": "7:21:00"} +{"current_steps": 3170, "total_steps": 7496, "loss": 0.4361, "lr": 0.0001429803406823134, "epoch": 3.383137673425827, "percentage": 42.29, "elapsed_time": "5:22:53", "remaining_time": "7:20:39"} +{"current_steps": 3175, "total_steps": 7496, "loss": 0.435, "lr": 0.00014276998016118683, "epoch": 3.3884738527214515, "percentage": 42.36, "elapsed_time": "5:23:53", "remaining_time": "7:20:47"} +{"current_steps": 3180, "total_steps": 7496, "loss": 0.4347, "lr": 0.0001425593877479712, "epoch": 3.393810032017076, "percentage": 42.42, "elapsed_time": "5:25:16", "remaining_time": "7:21:28"} +{"current_steps": 3185, "total_steps": 7496, "loss": 0.434, "lr": 0.0001423485645844655, "epoch": 3.3991462113127002, "percentage": 42.49, "elapsed_time": "5:26:45", "remaining_time": "7:22:17"} +{"current_steps": 3190, "total_steps": 7496, "loss": 0.4327, "lr": 0.0001421375118137196, "epoch": 3.4044823906083246, "percentage": 42.56, "elapsed_time": "5:27:21", "remaining_time": "7:21:52"} +{"current_steps": 3195, "total_steps": 7496, "loss": 0.4338, "lr": 0.00014192623058002833, "epoch": 3.409818569903949, "percentage": 42.62, "elapsed_time": "5:27:59", "remaining_time": "7:21:31"} +{"current_steps": 3200, "total_steps": 7496, "loss": 0.4329, "lr": 0.00014171472202892526, "epoch": 3.4151547491995733, "percentage": 42.69, "elapsed_time": "5:28:36", "remaining_time": "7:21:08"} +{"current_steps": 3205, "total_steps": 7496, "loss": 0.4352, "lr": 0.00014150298730717642, "epoch": 3.420490928495197, "percentage": 42.76, "elapsed_time": "5:29:23", "remaining_time": "7:21:00"} +{"current_steps": 3210, "total_steps": 7496, "loss": 0.4344, "lr": 0.00014129102756277404, "epoch": 3.4258271077908216, "percentage": 42.82, "elapsed_time": "5:30:01", "remaining_time": "7:20:39"} +{"current_steps": 3215, "total_steps": 7496, "loss": 0.433, "lr": 0.00014107884394493045, "epoch": 3.431163287086446, "percentage": 42.89, "elapsed_time": "5:30:39", "remaining_time": "7:20:17"} +{"current_steps": 3220, "total_steps": 7496, "loss": 0.4359, "lr": 0.00014086643760407184, "epoch": 3.4364994663820703, "percentage": 42.96, "elapsed_time": "5:31:14", "remaining_time": "7:19:52"} +{"current_steps": 3225, "total_steps": 7496, "loss": 0.4337, "lr": 0.00014065380969183178, "epoch": 3.4418356456776946, "percentage": 43.02, "elapsed_time": "5:31:51", "remaining_time": "7:19:30"} +{"current_steps": 3230, "total_steps": 7496, "loss": 0.4328, "lr": 0.00014044096136104538, "epoch": 3.447171824973319, "percentage": 43.09, "elapsed_time": "5:32:31", "remaining_time": "7:19:10"} +{"current_steps": 3235, "total_steps": 7496, "loss": 0.4323, "lr": 0.0001402278937657427, "epoch": 3.4525080042689433, "percentage": 43.16, "elapsed_time": "5:33:09", "remaining_time": "7:18:49"} +{"current_steps": 3240, "total_steps": 7496, "loss": 0.4347, "lr": 0.0001400146080611426, "epoch": 3.4578441835645677, "percentage": 43.22, "elapsed_time": "5:33:45", "remaining_time": "7:18:25"} +{"current_steps": 3245, "total_steps": 7496, "loss": 0.4326, "lr": 0.00013980110540364664, "epoch": 3.463180362860192, "percentage": 43.29, "elapsed_time": "5:34:23", "remaining_time": "7:18:04"} +{"current_steps": 3250, "total_steps": 7496, "loss": 0.433, "lr": 0.00013958738695083253, "epoch": 3.4685165421558164, "percentage": 43.36, "elapsed_time": "5:35:01", "remaining_time": "7:17:41"} +{"current_steps": 3255, "total_steps": 7496, "loss": 0.4322, "lr": 0.00013937345386144797, "epoch": 3.4738527214514408, "percentage": 43.42, "elapsed_time": "5:36:27", "remaining_time": "7:18:23"} +{"current_steps": 3260, "total_steps": 7496, "loss": 0.4336, "lr": 0.00013915930729540448, "epoch": 3.479188900747065, "percentage": 43.49, "elapsed_time": "5:38:08", "remaining_time": "7:19:23"} +{"current_steps": 3265, "total_steps": 7496, "loss": 0.4331, "lr": 0.00013894494841377102, "epoch": 3.4845250800426895, "percentage": 43.56, "elapsed_time": "5:38:44", "remaining_time": "7:18:58"} +{"current_steps": 3270, "total_steps": 7496, "loss": 0.4317, "lr": 0.00013873037837876758, "epoch": 3.489861259338314, "percentage": 43.62, "elapsed_time": "5:39:20", "remaining_time": "7:18:33"} +{"current_steps": 3275, "total_steps": 7496, "loss": 0.4353, "lr": 0.0001385155983537591, "epoch": 3.495197438633938, "percentage": 43.69, "elapsed_time": "5:39:59", "remaining_time": "7:18:11"} +{"current_steps": 3280, "total_steps": 7496, "loss": 0.4326, "lr": 0.00013830060950324898, "epoch": 3.5005336179295625, "percentage": 43.76, "elapsed_time": "5:40:38", "remaining_time": "7:17:50"} +{"current_steps": 3285, "total_steps": 7496, "loss": 0.4331, "lr": 0.0001380854129928729, "epoch": 3.505869797225187, "percentage": 43.82, "elapsed_time": "5:41:15", "remaining_time": "7:17:27"} +{"current_steps": 3290, "total_steps": 7496, "loss": 0.4346, "lr": 0.0001378700099893924, "epoch": 3.5112059765208112, "percentage": 43.89, "elapsed_time": "5:41:53", "remaining_time": "7:17:05"} +{"current_steps": 3295, "total_steps": 7496, "loss": 0.431, "lr": 0.0001376544016606886, "epoch": 3.5165421558164356, "percentage": 43.96, "elapsed_time": "5:43:05", "remaining_time": "7:17:25"} +{"current_steps": 3300, "total_steps": 7496, "loss": 0.4323, "lr": 0.00013743858917575588, "epoch": 3.52187833511206, "percentage": 44.02, "elapsed_time": "5:43:43", "remaining_time": "7:17:03"} +{"current_steps": 3305, "total_steps": 7496, "loss": 0.4316, "lr": 0.00013722257370469545, "epoch": 3.5272145144076843, "percentage": 44.09, "elapsed_time": "5:44:33", "remaining_time": "7:16:55"} +{"current_steps": 3310, "total_steps": 7496, "loss": 0.4337, "lr": 0.00013700635641870923, "epoch": 3.5325506937033087, "percentage": 44.16, "elapsed_time": "5:45:11", "remaining_time": "7:16:32"} +{"current_steps": 3315, "total_steps": 7496, "loss": 0.4316, "lr": 0.00013678993849009313, "epoch": 3.537886872998933, "percentage": 44.22, "elapsed_time": "5:45:46", "remaining_time": "7:16:06"} +{"current_steps": 3320, "total_steps": 7496, "loss": 0.434, "lr": 0.00013657332109223117, "epoch": 3.5432230522945574, "percentage": 44.29, "elapsed_time": "5:46:23", "remaining_time": "7:15:42"} +{"current_steps": 3325, "total_steps": 7496, "loss": 0.4328, "lr": 0.00013635650539958863, "epoch": 3.5485592315901813, "percentage": 44.36, "elapsed_time": "5:47:00", "remaining_time": "7:15:18"} +{"current_steps": 3330, "total_steps": 7496, "loss": 0.4326, "lr": 0.00013613949258770608, "epoch": 3.5538954108858056, "percentage": 44.42, "elapsed_time": "5:47:38", "remaining_time": "7:14:55"} +{"current_steps": 3335, "total_steps": 7496, "loss": 0.4332, "lr": 0.00013592228383319273, "epoch": 3.55923159018143, "percentage": 44.49, "elapsed_time": "5:48:14", "remaining_time": "7:14:29"} +{"current_steps": 3340, "total_steps": 7496, "loss": 0.4331, "lr": 0.00013570488031372022, "epoch": 3.5645677694770543, "percentage": 44.56, "elapsed_time": "5:48:52", "remaining_time": "7:14:06"} +{"current_steps": 3345, "total_steps": 7496, "loss": 0.4338, "lr": 0.00013548728320801612, "epoch": 3.5699039487726787, "percentage": 44.62, "elapsed_time": "5:49:28", "remaining_time": "7:13:41"} +{"current_steps": 3350, "total_steps": 7496, "loss": 0.4314, "lr": 0.00013526949369585764, "epoch": 3.575240128068303, "percentage": 44.69, "elapsed_time": "5:50:08", "remaining_time": "7:13:20"} +{"current_steps": 3355, "total_steps": 7496, "loss": 0.4318, "lr": 0.0001350515129580652, "epoch": 3.5805763073639274, "percentage": 44.76, "elapsed_time": "5:50:43", "remaining_time": "7:12:53"} +{"current_steps": 3360, "total_steps": 7496, "loss": 0.4337, "lr": 0.000134833342176496, "epoch": 3.5859124866595518, "percentage": 44.82, "elapsed_time": "5:51:19", "remaining_time": "7:12:27"} +{"current_steps": 3365, "total_steps": 7496, "loss": 0.4337, "lr": 0.00013461498253403758, "epoch": 3.591248665955176, "percentage": 44.89, "elapsed_time": "5:51:59", "remaining_time": "7:12:07"} +{"current_steps": 3370, "total_steps": 7496, "loss": 0.43, "lr": 0.00013439643521460153, "epoch": 3.5965848452508005, "percentage": 44.96, "elapsed_time": "5:52:35", "remaining_time": "7:11:41"} +{"current_steps": 3375, "total_steps": 7496, "loss": 0.4315, "lr": 0.00013417770140311693, "epoch": 3.601921024546425, "percentage": 45.02, "elapsed_time": "5:53:11", "remaining_time": "7:11:15"} +{"current_steps": 3380, "total_steps": 7496, "loss": 0.4304, "lr": 0.000133958782285524, "epoch": 3.607257203842049, "percentage": 45.09, "elapsed_time": "5:53:49", "remaining_time": "7:10:52"} +{"current_steps": 3385, "total_steps": 7496, "loss": 0.4337, "lr": 0.00013373967904876775, "epoch": 3.6125933831376735, "percentage": 45.16, "elapsed_time": "5:54:26", "remaining_time": "7:10:27"} +{"current_steps": 3390, "total_steps": 7496, "loss": 0.4306, "lr": 0.00013352039288079133, "epoch": 3.617929562433298, "percentage": 45.22, "elapsed_time": "5:55:01", "remaining_time": "7:10:01"} +{"current_steps": 3395, "total_steps": 7496, "loss": 0.433, "lr": 0.00013330092497052976, "epoch": 3.6232657417289222, "percentage": 45.29, "elapsed_time": "5:55:39", "remaining_time": "7:09:37"} +{"current_steps": 3400, "total_steps": 7496, "loss": 0.4319, "lr": 0.00013308127650790352, "epoch": 3.628601921024546, "percentage": 45.36, "elapsed_time": "5:56:18", "remaining_time": "7:09:15"} +{"current_steps": 3405, "total_steps": 7496, "loss": 0.4325, "lr": 0.00013286144868381188, "epoch": 3.6339381003201705, "percentage": 45.42, "elapsed_time": "5:57:04", "remaining_time": "7:09:00"} +{"current_steps": 3410, "total_steps": 7496, "loss": 0.4321, "lr": 0.0001326414426901266, "epoch": 3.639274279615795, "percentage": 45.49, "elapsed_time": "5:57:43", "remaining_time": "7:08:38"} +{"current_steps": 3415, "total_steps": 7496, "loss": 0.4321, "lr": 0.00013242125971968556, "epoch": 3.644610458911419, "percentage": 45.56, "elapsed_time": "5:58:21", "remaining_time": "7:08:15"} +{"current_steps": 3420, "total_steps": 7496, "loss": 0.4336, "lr": 0.00013220090096628616, "epoch": 3.6499466382070436, "percentage": 45.62, "elapsed_time": "5:58:56", "remaining_time": "7:07:47"} +{"current_steps": 3425, "total_steps": 7496, "loss": 0.432, "lr": 0.00013198036762467868, "epoch": 3.655282817502668, "percentage": 45.69, "elapsed_time": "5:59:35", "remaining_time": "7:07:24"} +{"current_steps": 3430, "total_steps": 7496, "loss": 0.4316, "lr": 0.00013175966089056025, "epoch": 3.6606189967982923, "percentage": 45.76, "elapsed_time": "6:00:15", "remaining_time": "7:07:03"} +{"current_steps": 3435, "total_steps": 7496, "loss": 0.4304, "lr": 0.00013153878196056788, "epoch": 3.6659551760939166, "percentage": 45.82, "elapsed_time": "6:00:52", "remaining_time": "7:06:38"} +{"current_steps": 3440, "total_steps": 7496, "loss": 0.4304, "lr": 0.00013131773203227238, "epoch": 3.671291355389541, "percentage": 45.89, "elapsed_time": "6:01:27", "remaining_time": "7:06:11"} +{"current_steps": 3445, "total_steps": 7496, "loss": 0.4295, "lr": 0.00013109651230417158, "epoch": 3.6766275346851653, "percentage": 45.96, "elapsed_time": "6:02:07", "remaining_time": "7:05:49"} +{"current_steps": 3450, "total_steps": 7496, "loss": 0.4321, "lr": 0.000130875123975684, "epoch": 3.6819637139807897, "percentage": 46.02, "elapsed_time": "6:02:44", "remaining_time": "7:05:24"} +{"current_steps": 3455, "total_steps": 7496, "loss": 0.4318, "lr": 0.00013065356824714218, "epoch": 3.687299893276414, "percentage": 46.09, "elapsed_time": "6:03:19", "remaining_time": "7:04:57"} +{"current_steps": 3460, "total_steps": 7496, "loss": 0.4303, "lr": 0.0001304318463197864, "epoch": 3.6926360725720384, "percentage": 46.16, "elapsed_time": "6:03:57", "remaining_time": "7:04:33"} +{"current_steps": 3465, "total_steps": 7496, "loss": 0.4293, "lr": 0.00013020995939575801, "epoch": 3.6979722518676628, "percentage": 46.22, "elapsed_time": "6:04:36", "remaining_time": "7:04:09"} +{"current_steps": 3470, "total_steps": 7496, "loss": 0.4286, "lr": 0.0001299879086780929, "epoch": 3.703308431163287, "percentage": 46.29, "elapsed_time": "6:05:12", "remaining_time": "7:03:43"} +{"current_steps": 3475, "total_steps": 7496, "loss": 0.4305, "lr": 0.000129765695370715, "epoch": 3.7086446104589115, "percentage": 46.36, "elapsed_time": "6:05:51", "remaining_time": "7:03:20"} +{"current_steps": 3480, "total_steps": 7496, "loss": 0.4308, "lr": 0.00012954332067842994, "epoch": 3.713980789754536, "percentage": 46.42, "elapsed_time": "6:06:31", "remaining_time": "7:02:58"} +{"current_steps": 3485, "total_steps": 7496, "loss": 0.4325, "lr": 0.00012932078580691817, "epoch": 3.71931696905016, "percentage": 46.49, "elapsed_time": "6:07:07", "remaining_time": "7:02:32"} +{"current_steps": 3490, "total_steps": 7496, "loss": 0.4289, "lr": 0.00012909809196272865, "epoch": 3.7246531483457845, "percentage": 46.56, "elapsed_time": "6:07:44", "remaining_time": "7:02:07"} +{"current_steps": 3495, "total_steps": 7496, "loss": 0.4309, "lr": 0.00012887524035327233, "epoch": 3.729989327641409, "percentage": 46.62, "elapsed_time": "6:08:24", "remaining_time": "7:01:44"} +{"current_steps": 3500, "total_steps": 7496, "loss": 0.4316, "lr": 0.00012865223218681555, "epoch": 3.7353255069370332, "percentage": 46.69, "elapsed_time": "6:08:58", "remaining_time": "7:01:16"} +{"current_steps": 3505, "total_steps": 7496, "loss": 0.4278, "lr": 0.0001284290686724733, "epoch": 3.7406616862326576, "percentage": 46.76, "elapsed_time": "6:09:46", "remaining_time": "7:01:02"} +{"current_steps": 3510, "total_steps": 7496, "loss": 0.4291, "lr": 0.00012820575102020308, "epoch": 3.745997865528282, "percentage": 46.82, "elapsed_time": "6:10:23", "remaining_time": "7:00:37"} +{"current_steps": 3515, "total_steps": 7496, "loss": 0.4301, "lr": 0.00012798228044079795, "epoch": 3.7513340448239063, "percentage": 46.89, "elapsed_time": "6:11:01", "remaining_time": "7:00:13"} +{"current_steps": 3520, "total_steps": 7496, "loss": 0.4281, "lr": 0.0001277586581458802, "epoch": 3.7566702241195307, "percentage": 46.96, "elapsed_time": "6:11:38", "remaining_time": "6:59:47"} +{"current_steps": 3525, "total_steps": 7496, "loss": 0.429, "lr": 0.00012753488534789458, "epoch": 3.762006403415155, "percentage": 47.03, "elapsed_time": "6:12:14", "remaining_time": "6:59:20"} +{"current_steps": 3530, "total_steps": 7496, "loss": 0.4314, "lr": 0.00012731096326010194, "epoch": 3.767342582710779, "percentage": 47.09, "elapsed_time": "6:12:53", "remaining_time": "6:58:56"} +{"current_steps": 3535, "total_steps": 7496, "loss": 0.4322, "lr": 0.00012708689309657255, "epoch": 3.7726787620064033, "percentage": 47.16, "elapsed_time": "6:13:28", "remaining_time": "6:58:29"} +{"current_steps": 3540, "total_steps": 7496, "loss": 0.4277, "lr": 0.00012686267607217954, "epoch": 3.7780149413020276, "percentage": 47.23, "elapsed_time": "6:14:04", "remaining_time": "6:58:02"} +{"current_steps": 3545, "total_steps": 7496, "loss": 0.4307, "lr": 0.00012663831340259222, "epoch": 3.783351120597652, "percentage": 47.29, "elapsed_time": "6:14:41", "remaining_time": "6:57:36"} +{"current_steps": 3550, "total_steps": 7496, "loss": 0.4307, "lr": 0.00012641380630426963, "epoch": 3.7886872998932764, "percentage": 47.36, "elapsed_time": "6:15:18", "remaining_time": "6:57:10"} +{"current_steps": 3555, "total_steps": 7496, "loss": 0.4282, "lr": 0.00012618915599445387, "epoch": 3.7940234791889007, "percentage": 47.43, "elapsed_time": "6:15:54", "remaining_time": "6:56:43"} +{"current_steps": 3560, "total_steps": 7496, "loss": 0.4291, "lr": 0.00012596436369116348, "epoch": 3.799359658484525, "percentage": 47.49, "elapsed_time": "6:16:31", "remaining_time": "6:56:17"} +{"current_steps": 3565, "total_steps": 7496, "loss": 0.4276, "lr": 0.00012573943061318686, "epoch": 3.8046958377801494, "percentage": 47.56, "elapsed_time": "6:17:10", "remaining_time": "6:55:53"} +{"current_steps": 3570, "total_steps": 7496, "loss": 0.4285, "lr": 0.00012551435798007572, "epoch": 3.8100320170757738, "percentage": 47.63, "elapsed_time": "6:17:47", "remaining_time": "6:55:27"} +{"current_steps": 3575, "total_steps": 7496, "loss": 0.4298, "lr": 0.0001252891470121384, "epoch": 3.815368196371398, "percentage": 47.69, "elapsed_time": "6:18:22", "remaining_time": "6:54:59"} +{"current_steps": 3580, "total_steps": 7496, "loss": 0.4302, "lr": 0.00012506379893043328, "epoch": 3.8207043756670225, "percentage": 47.76, "elapsed_time": "6:19:00", "remaining_time": "6:54:34"} +{"current_steps": 3585, "total_steps": 7496, "loss": 0.4294, "lr": 0.00012483831495676206, "epoch": 3.826040554962647, "percentage": 47.83, "elapsed_time": "6:19:35", "remaining_time": "6:54:07"} +{"current_steps": 3590, "total_steps": 7496, "loss": 0.4294, "lr": 0.00012461269631366338, "epoch": 3.831376734258271, "percentage": 47.89, "elapsed_time": "6:20:13", "remaining_time": "6:53:41"} +{"current_steps": 3595, "total_steps": 7496, "loss": 0.4293, "lr": 0.00012438694422440587, "epoch": 3.8367129135538955, "percentage": 47.96, "elapsed_time": "6:20:49", "remaining_time": "6:53:14"} +{"current_steps": 3600, "total_steps": 7496, "loss": 0.429, "lr": 0.0001241610599129818, "epoch": 3.84204909284952, "percentage": 48.03, "elapsed_time": "6:21:25", "remaining_time": "6:52:47"} +{"current_steps": 3605, "total_steps": 7496, "loss": 0.4291, "lr": 0.0001239350446041003, "epoch": 3.847385272145144, "percentage": 48.09, "elapsed_time": "6:22:15", "remaining_time": "6:52:34"} +{"current_steps": 3610, "total_steps": 7496, "loss": 0.4289, "lr": 0.00012370889952318076, "epoch": 3.852721451440768, "percentage": 48.16, "elapsed_time": "6:22:51", "remaining_time": "6:52:07"} +{"current_steps": 3615, "total_steps": 7496, "loss": 0.4289, "lr": 0.0001234826258963461, "epoch": 3.8580576307363925, "percentage": 48.23, "elapsed_time": "6:23:30", "remaining_time": "6:51:43"} +{"current_steps": 3620, "total_steps": 7496, "loss": 0.4304, "lr": 0.00012325622495041626, "epoch": 3.863393810032017, "percentage": 48.29, "elapsed_time": "6:24:07", "remaining_time": "6:51:17"} +{"current_steps": 3625, "total_steps": 7496, "loss": 0.4273, "lr": 0.00012302969791290152, "epoch": 3.8687299893276412, "percentage": 48.36, "elapsed_time": "6:24:44", "remaining_time": "6:50:50"} +{"current_steps": 3630, "total_steps": 7496, "loss": 0.43, "lr": 0.00012280304601199567, "epoch": 3.8740661686232656, "percentage": 48.43, "elapsed_time": "6:25:22", "remaining_time": "6:50:26"} +{"current_steps": 3635, "total_steps": 7496, "loss": 0.4295, "lr": 0.0001225762704765697, "epoch": 3.87940234791889, "percentage": 48.49, "elapsed_time": "6:25:58", "remaining_time": "6:49:57"} +{"current_steps": 3640, "total_steps": 7496, "loss": 0.4283, "lr": 0.0001223493725361647, "epoch": 3.8847385272145143, "percentage": 48.56, "elapsed_time": "6:26:36", "remaining_time": "6:49:33"} +{"current_steps": 3645, "total_steps": 7496, "loss": 0.429, "lr": 0.00012212235342098563, "epoch": 3.8900747065101386, "percentage": 48.63, "elapsed_time": "6:27:15", "remaining_time": "6:49:08"} +{"current_steps": 3650, "total_steps": 7496, "loss": 0.4274, "lr": 0.00012189521436189427, "epoch": 3.895410885805763, "percentage": 48.69, "elapsed_time": "6:27:51", "remaining_time": "6:48:40"} +{"current_steps": 3655, "total_steps": 7496, "loss": 0.429, "lr": 0.00012166795659040278, "epoch": 3.9007470651013874, "percentage": 48.76, "elapsed_time": "6:28:28", "remaining_time": "6:48:14"} +{"current_steps": 3660, "total_steps": 7496, "loss": 0.4281, "lr": 0.00012144058133866696, "epoch": 3.9060832443970117, "percentage": 48.83, "elapsed_time": "6:29:03", "remaining_time": "6:47:46"} +{"current_steps": 3665, "total_steps": 7496, "loss": 0.4296, "lr": 0.00012121308983947963, "epoch": 3.911419423692636, "percentage": 48.89, "elapsed_time": "6:29:40", "remaining_time": "6:47:19"} +{"current_steps": 3670, "total_steps": 7496, "loss": 0.4278, "lr": 0.00012098548332626383, "epoch": 3.9167556029882604, "percentage": 48.96, "elapsed_time": "6:30:16", "remaining_time": "6:46:52"} +{"current_steps": 3675, "total_steps": 7496, "loss": 0.428, "lr": 0.00012075776303306612, "epoch": 3.9220917822838848, "percentage": 49.03, "elapsed_time": "6:30:54", "remaining_time": "6:46:26"} +{"current_steps": 3680, "total_steps": 7496, "loss": 0.4252, "lr": 0.00012052993019455007, "epoch": 3.927427961579509, "percentage": 49.09, "elapsed_time": "6:31:31", "remaining_time": "6:45:59"} +{"current_steps": 3685, "total_steps": 7496, "loss": 0.4267, "lr": 0.00012030198604598943, "epoch": 3.9327641408751335, "percentage": 49.16, "elapsed_time": "6:32:06", "remaining_time": "6:45:30"} +{"current_steps": 3690, "total_steps": 7496, "loss": 0.4268, "lr": 0.00012007393182326136, "epoch": 3.938100320170758, "percentage": 49.23, "elapsed_time": "6:32:44", "remaining_time": "6:45:05"} +{"current_steps": 3695, "total_steps": 7496, "loss": 0.4276, "lr": 0.00011984576876284, "epoch": 3.943436499466382, "percentage": 49.29, "elapsed_time": "6:33:21", "remaining_time": "6:44:38"} +{"current_steps": 3700, "total_steps": 7496, "loss": 0.4272, "lr": 0.00011961749810178943, "epoch": 3.9487726787620065, "percentage": 49.36, "elapsed_time": "6:33:57", "remaining_time": "6:44:10"} +{"current_steps": 3705, "total_steps": 7496, "loss": 0.4273, "lr": 0.00011938912107775717, "epoch": 3.954108858057631, "percentage": 49.43, "elapsed_time": "6:34:47", "remaining_time": "6:43:57"} +{"current_steps": 3710, "total_steps": 7496, "loss": 0.428, "lr": 0.00011916063892896748, "epoch": 3.9594450373532553, "percentage": 49.49, "elapsed_time": "6:35:25", "remaining_time": "6:43:31"} +{"current_steps": 3715, "total_steps": 7496, "loss": 0.4277, "lr": 0.00011893205289421451, "epoch": 3.9647812166488796, "percentage": 49.56, "elapsed_time": "6:35:59", "remaining_time": "6:43:01"} +{"current_steps": 3720, "total_steps": 7496, "loss": 0.4286, "lr": 0.00011870336421285571, "epoch": 3.970117395944504, "percentage": 49.63, "elapsed_time": "6:36:36", "remaining_time": "6:42:34"} +{"current_steps": 3725, "total_steps": 7496, "loss": 0.426, "lr": 0.00011847457412480505, "epoch": 3.9754535752401283, "percentage": 49.69, "elapsed_time": "6:37:11", "remaining_time": "6:42:06"} +{"current_steps": 3730, "total_steps": 7496, "loss": 0.4288, "lr": 0.00011824568387052637, "epoch": 3.9807897545357527, "percentage": 49.76, "elapsed_time": "6:37:47", "remaining_time": "6:41:38"} +{"current_steps": 3735, "total_steps": 7496, "loss": 0.427, "lr": 0.00011801669469102643, "epoch": 3.9861259338313766, "percentage": 49.83, "elapsed_time": "6:38:23", "remaining_time": "6:41:09"} +{"current_steps": 3740, "total_steps": 7496, "loss": 0.4273, "lr": 0.00011778760782784855, "epoch": 3.991462113127001, "percentage": 49.89, "elapsed_time": "6:38:57", "remaining_time": "6:40:40"} +{"current_steps": 3745, "total_steps": 7496, "loss": 0.4273, "lr": 0.00011755842452306556, "epoch": 3.9967982924226253, "percentage": 49.96, "elapsed_time": "6:39:32", "remaining_time": "6:40:11"} +{"current_steps": 3750, "total_steps": 7496, "loss": 0.4239, "lr": 0.00011732914601927318, "epoch": 4.00213447171825, "percentage": 50.03, "elapsed_time": "6:40:10", "remaining_time": "6:39:44"} +{"current_steps": 3755, "total_steps": 7496, "loss": 0.4218, "lr": 0.00011709977355958333, "epoch": 4.0074706510138745, "percentage": 50.09, "elapsed_time": "6:40:44", "remaining_time": "6:39:15"} +{"current_steps": 3760, "total_steps": 7496, "loss": 0.4232, "lr": 0.00011687030838761732, "epoch": 4.012806830309499, "percentage": 50.16, "elapsed_time": "6:41:18", "remaining_time": "6:38:45"} +{"current_steps": 3765, "total_steps": 7496, "loss": 0.4231, "lr": 0.00011664075174749911, "epoch": 4.018143009605123, "percentage": 50.23, "elapsed_time": "6:41:55", "remaining_time": "6:38:17"} +{"current_steps": 3770, "total_steps": 7496, "loss": 0.4239, "lr": 0.00011641110488384866, "epoch": 4.023479188900747, "percentage": 50.29, "elapsed_time": "6:42:33", "remaining_time": "6:37:51"} +{"current_steps": 3775, "total_steps": 7496, "loss": 0.4225, "lr": 0.00011618136904177501, "epoch": 4.028815368196371, "percentage": 50.36, "elapsed_time": "6:43:07", "remaining_time": "6:37:21"} +{"current_steps": 3780, "total_steps": 7496, "loss": 0.4236, "lr": 0.00011595154546686969, "epoch": 4.034151547491995, "percentage": 50.43, "elapsed_time": "6:43:42", "remaining_time": "6:36:52"} +{"current_steps": 3785, "total_steps": 7496, "loss": 0.4228, "lr": 0.00011572163540519986, "epoch": 4.03948772678762, "percentage": 50.49, "elapsed_time": "6:44:22", "remaining_time": "6:36:27"} +{"current_steps": 3790, "total_steps": 7496, "loss": 0.4227, "lr": 0.00011549164010330171, "epoch": 4.044823906083244, "percentage": 50.56, "elapsed_time": "6:44:57", "remaining_time": "6:35:59"} +{"current_steps": 3795, "total_steps": 7496, "loss": 0.4239, "lr": 0.00011526156080817341, "epoch": 4.050160085378868, "percentage": 50.63, "elapsed_time": "6:45:34", "remaining_time": "6:35:31"} +{"current_steps": 3800, "total_steps": 7496, "loss": 0.4248, "lr": 0.00011503139876726863, "epoch": 4.055496264674493, "percentage": 50.69, "elapsed_time": "6:46:34", "remaining_time": "6:35:26"} +{"current_steps": 3805, "total_steps": 7496, "loss": 0.4244, "lr": 0.00011480115522848973, "epoch": 4.060832443970117, "percentage": 50.76, "elapsed_time": "6:47:45", "remaining_time": "6:35:32"} +{"current_steps": 3810, "total_steps": 7496, "loss": 0.422, "lr": 0.00011457083144018083, "epoch": 4.0661686232657415, "percentage": 50.83, "elapsed_time": "6:48:49", "remaining_time": "6:35:30"} +{"current_steps": 3815, "total_steps": 7496, "loss": 0.4231, "lr": 0.0001143404286511212, "epoch": 4.071504802561366, "percentage": 50.89, "elapsed_time": "6:50:32", "remaining_time": "6:36:07"} +{"current_steps": 3820, "total_steps": 7496, "loss": 0.4214, "lr": 0.00011410994811051847, "epoch": 4.07684098185699, "percentage": 50.96, "elapsed_time": "6:51:33", "remaining_time": "6:36:02"} +{"current_steps": 3825, "total_steps": 7496, "loss": 0.4213, "lr": 0.00011387939106800176, "epoch": 4.0821771611526145, "percentage": 51.03, "elapsed_time": "6:52:08", "remaining_time": "6:35:33"} +{"current_steps": 3830, "total_steps": 7496, "loss": 0.4225, "lr": 0.00011364875877361496, "epoch": 4.087513340448239, "percentage": 51.09, "elapsed_time": "6:53:13", "remaining_time": "6:35:31"} +{"current_steps": 3835, "total_steps": 7496, "loss": 0.4204, "lr": 0.00011341805247781008, "epoch": 4.092849519743863, "percentage": 51.16, "elapsed_time": "6:53:48", "remaining_time": "6:35:02"} +{"current_steps": 3840, "total_steps": 7496, "loss": 0.4233, "lr": 0.00011318727343144026, "epoch": 4.098185699039488, "percentage": 51.23, "elapsed_time": "6:54:24", "remaining_time": "6:34:33"} +{"current_steps": 3845, "total_steps": 7496, "loss": 0.4219, "lr": 0.00011295642288575307, "epoch": 4.103521878335112, "percentage": 51.29, "elapsed_time": "6:55:01", "remaining_time": "6:34:05"} +{"current_steps": 3850, "total_steps": 7496, "loss": 0.4215, "lr": 0.00011272550209238387, "epoch": 4.108858057630736, "percentage": 51.36, "elapsed_time": "6:55:37", "remaining_time": "6:33:36"} +{"current_steps": 3855, "total_steps": 7496, "loss": 0.4213, "lr": 0.00011249451230334867, "epoch": 4.114194236926361, "percentage": 51.43, "elapsed_time": "6:56:14", "remaining_time": "6:33:07"} +{"current_steps": 3860, "total_steps": 7496, "loss": 0.4215, "lr": 0.00011226345477103773, "epoch": 4.119530416221985, "percentage": 51.49, "elapsed_time": "6:56:51", "remaining_time": "6:32:40"} +{"current_steps": 3865, "total_steps": 7496, "loss": 0.4216, "lr": 0.00011203233074820858, "epoch": 4.124866595517609, "percentage": 51.56, "elapsed_time": "6:57:26", "remaining_time": "6:32:10"} +{"current_steps": 3870, "total_steps": 7496, "loss": 0.4237, "lr": 0.00011180114148797922, "epoch": 4.130202774813234, "percentage": 51.63, "elapsed_time": "6:58:01", "remaining_time": "6:31:40"} +{"current_steps": 3875, "total_steps": 7496, "loss": 0.4228, "lr": 0.00011156988824382135, "epoch": 4.135538954108858, "percentage": 51.69, "elapsed_time": "6:58:38", "remaining_time": "6:31:12"} +{"current_steps": 3880, "total_steps": 7496, "loss": 0.4209, "lr": 0.00011133857226955364, "epoch": 4.140875133404482, "percentage": 51.76, "elapsed_time": "6:59:16", "remaining_time": "6:30:44"} +{"current_steps": 3885, "total_steps": 7496, "loss": 0.4235, "lr": 0.00011110719481933481, "epoch": 4.146211312700107, "percentage": 51.83, "elapsed_time": "6:59:52", "remaining_time": "6:30:15"} +{"current_steps": 3890, "total_steps": 7496, "loss": 0.4234, "lr": 0.00011087575714765688, "epoch": 4.151547491995731, "percentage": 51.89, "elapsed_time": "7:00:29", "remaining_time": "6:29:47"} +{"current_steps": 3895, "total_steps": 7496, "loss": 0.4217, "lr": 0.00011064426050933847, "epoch": 4.1568836712913555, "percentage": 51.96, "elapsed_time": "7:01:06", "remaining_time": "6:29:19"} +{"current_steps": 3900, "total_steps": 7496, "loss": 0.4223, "lr": 0.00011041270615951782, "epoch": 4.16221985058698, "percentage": 52.03, "elapsed_time": "7:01:43", "remaining_time": "6:28:51"} +{"current_steps": 3905, "total_steps": 7496, "loss": 0.421, "lr": 0.00011018109535364613, "epoch": 4.167556029882604, "percentage": 52.09, "elapsed_time": "7:02:30", "remaining_time": "6:28:32"} +{"current_steps": 3910, "total_steps": 7496, "loss": 0.4211, "lr": 0.00010994942934748065, "epoch": 4.172892209178229, "percentage": 52.16, "elapsed_time": "7:03:08", "remaining_time": "6:28:04"} +{"current_steps": 3915, "total_steps": 7496, "loss": 0.422, "lr": 0.00010971770939707795, "epoch": 4.178228388473853, "percentage": 52.23, "elapsed_time": "7:04:08", "remaining_time": "6:27:57"} +{"current_steps": 3920, "total_steps": 7496, "loss": 0.4221, "lr": 0.00010948593675878706, "epoch": 4.183564567769477, "percentage": 52.29, "elapsed_time": "7:04:40", "remaining_time": "6:27:24"} +{"current_steps": 3925, "total_steps": 7496, "loss": 0.4205, "lr": 0.00010925411268924268, "epoch": 4.188900747065102, "percentage": 52.36, "elapsed_time": "7:05:18", "remaining_time": "6:26:57"} +{"current_steps": 3930, "total_steps": 7496, "loss": 0.4224, "lr": 0.00010902223844535839, "epoch": 4.194236926360726, "percentage": 52.43, "elapsed_time": "7:05:54", "remaining_time": "6:26:27"} +{"current_steps": 3935, "total_steps": 7496, "loss": 0.42, "lr": 0.00010879031528431975, "epoch": 4.19957310565635, "percentage": 52.49, "elapsed_time": "7:06:29", "remaining_time": "6:25:57"} +{"current_steps": 3940, "total_steps": 7496, "loss": 0.423, "lr": 0.0001085583444635776, "epoch": 4.204909284951975, "percentage": 52.56, "elapsed_time": "7:07:06", "remaining_time": "6:25:28"} +{"current_steps": 3945, "total_steps": 7496, "loss": 0.4217, "lr": 0.00010832632724084113, "epoch": 4.210245464247599, "percentage": 52.63, "elapsed_time": "7:07:42", "remaining_time": "6:24:59"} +{"current_steps": 3950, "total_steps": 7496, "loss": 0.419, "lr": 0.00010809426487407115, "epoch": 4.215581643543223, "percentage": 52.69, "elapsed_time": "7:08:16", "remaining_time": "6:24:28"} +{"current_steps": 3955, "total_steps": 7496, "loss": 0.4225, "lr": 0.0001078621586214732, "epoch": 4.220917822838848, "percentage": 52.76, "elapsed_time": "7:08:52", "remaining_time": "6:23:59"} +{"current_steps": 3960, "total_steps": 7496, "loss": 0.4213, "lr": 0.00010763000974149084, "epoch": 4.226254002134472, "percentage": 52.83, "elapsed_time": "7:09:29", "remaining_time": "6:23:29"} +{"current_steps": 3965, "total_steps": 7496, "loss": 0.4204, "lr": 0.00010739781949279865, "epoch": 4.2315901814300965, "percentage": 52.89, "elapsed_time": "7:10:05", "remaining_time": "6:23:00"} +{"current_steps": 3970, "total_steps": 7496, "loss": 0.4222, "lr": 0.00010716558913429557, "epoch": 4.23692636072572, "percentage": 52.96, "elapsed_time": "7:10:37", "remaining_time": "6:22:28"} +{"current_steps": 3975, "total_steps": 7496, "loss": 0.4219, "lr": 0.00010693331992509801, "epoch": 4.242262540021345, "percentage": 53.03, "elapsed_time": "7:11:50", "remaining_time": "6:22:30"} +{"current_steps": 3980, "total_steps": 7496, "loss": 0.4215, "lr": 0.00010670101312453293, "epoch": 4.247598719316969, "percentage": 53.09, "elapsed_time": "7:12:26", "remaining_time": "6:22:01"} +{"current_steps": 3985, "total_steps": 7496, "loss": 0.4203, "lr": 0.00010646866999213123, "epoch": 4.252934898612593, "percentage": 53.16, "elapsed_time": "7:13:00", "remaining_time": "6:21:29"} +{"current_steps": 3990, "total_steps": 7496, "loss": 0.4218, "lr": 0.00010623629178762073, "epoch": 4.258271077908217, "percentage": 53.23, "elapsed_time": "7:13:37", "remaining_time": "6:21:01"} +{"current_steps": 3995, "total_steps": 7496, "loss": 0.4217, "lr": 0.00010600387977091944, "epoch": 4.263607257203842, "percentage": 53.3, "elapsed_time": "7:14:14", "remaining_time": "6:20:33"} +{"current_steps": 4000, "total_steps": 7496, "loss": 0.4238, "lr": 0.00010577143520212865, "epoch": 4.268943436499466, "percentage": 53.36, "elapsed_time": "7:14:47", "remaining_time": "6:20:00"} +{"current_steps": 4005, "total_steps": 7496, "loss": 0.4208, "lr": 0.00010553895934152613, "epoch": 4.27427961579509, "percentage": 53.43, "elapsed_time": "7:15:35", "remaining_time": "6:19:40"} +{"current_steps": 4010, "total_steps": 7496, "loss": 0.4204, "lr": 0.0001053064534495594, "epoch": 4.279615795090715, "percentage": 53.5, "elapsed_time": "7:16:11", "remaining_time": "6:19:11"} +{"current_steps": 4015, "total_steps": 7496, "loss": 0.4215, "lr": 0.00010507391878683867, "epoch": 4.284951974386339, "percentage": 53.56, "elapsed_time": "7:16:44", "remaining_time": "6:18:39"} +{"current_steps": 4020, "total_steps": 7496, "loss": 0.4209, "lr": 0.00010484135661413032, "epoch": 4.2902881536819635, "percentage": 53.63, "elapsed_time": "7:17:18", "remaining_time": "6:18:08"} +{"current_steps": 4025, "total_steps": 7496, "loss": 0.4221, "lr": 0.00010460876819234972, "epoch": 4.295624332977588, "percentage": 53.7, "elapsed_time": "7:17:56", "remaining_time": "6:17:40"} +{"current_steps": 4030, "total_steps": 7496, "loss": 0.4199, "lr": 0.00010437615478255464, "epoch": 4.300960512273212, "percentage": 53.76, "elapsed_time": "7:18:31", "remaining_time": "6:17:09"} +{"current_steps": 4035, "total_steps": 7496, "loss": 0.4206, "lr": 0.0001041435176459383, "epoch": 4.3062966915688365, "percentage": 53.83, "elapsed_time": "7:19:05", "remaining_time": "6:16:38"} +{"current_steps": 4040, "total_steps": 7496, "loss": 0.4193, "lr": 0.00010391085804382258, "epoch": 4.311632870864461, "percentage": 53.9, "elapsed_time": "7:19:43", "remaining_time": "6:16:09"} +{"current_steps": 4045, "total_steps": 7496, "loss": 0.4178, "lr": 0.00010367817723765118, "epoch": 4.316969050160085, "percentage": 53.96, "elapsed_time": "7:20:19", "remaining_time": "6:15:39"} +{"current_steps": 4050, "total_steps": 7496, "loss": 0.4212, "lr": 0.00010344547648898268, "epoch": 4.32230522945571, "percentage": 54.03, "elapsed_time": "7:20:54", "remaining_time": "6:15:09"} +{"current_steps": 4055, "total_steps": 7496, "loss": 0.4208, "lr": 0.00010321275705948395, "epoch": 4.327641408751334, "percentage": 54.1, "elapsed_time": "7:21:30", "remaining_time": "6:14:39"} +{"current_steps": 4060, "total_steps": 7496, "loss": 0.4211, "lr": 0.00010298002021092295, "epoch": 4.332977588046958, "percentage": 54.16, "elapsed_time": "7:22:06", "remaining_time": "6:14:09"} +{"current_steps": 4065, "total_steps": 7496, "loss": 0.4222, "lr": 0.00010274726720516224, "epoch": 4.338313767342583, "percentage": 54.23, "elapsed_time": "7:22:44", "remaining_time": "6:13:41"} +{"current_steps": 4070, "total_steps": 7496, "loss": 0.4211, "lr": 0.00010251449930415189, "epoch": 4.343649946638207, "percentage": 54.3, "elapsed_time": "7:23:17", "remaining_time": "6:13:09"} +{"current_steps": 4075, "total_steps": 7496, "loss": 0.4198, "lr": 0.00010228171776992277, "epoch": 4.348986125933831, "percentage": 54.36, "elapsed_time": "7:23:53", "remaining_time": "6:12:39"} +{"current_steps": 4080, "total_steps": 7496, "loss": 0.419, "lr": 0.00010204892386457963, "epoch": 4.354322305229456, "percentage": 54.43, "elapsed_time": "7:24:30", "remaining_time": "6:12:10"} +{"current_steps": 4085, "total_steps": 7496, "loss": 0.4206, "lr": 0.00010181611885029437, "epoch": 4.35965848452508, "percentage": 54.5, "elapsed_time": "7:25:08", "remaining_time": "6:11:41"} +{"current_steps": 4090, "total_steps": 7496, "loss": 0.4217, "lr": 0.00010158330398929907, "epoch": 4.3649946638207044, "percentage": 54.56, "elapsed_time": "7:25:42", "remaining_time": "6:11:10"} +{"current_steps": 4095, "total_steps": 7496, "loss": 0.4208, "lr": 0.00010135048054387913, "epoch": 4.370330843116329, "percentage": 54.63, "elapsed_time": "7:26:56", "remaining_time": "6:11:11"} +{"current_steps": 4100, "total_steps": 7496, "loss": 0.4198, "lr": 0.00010111764977636662, "epoch": 4.375667022411953, "percentage": 54.7, "elapsed_time": "7:27:32", "remaining_time": "6:10:41"} +{"current_steps": 4105, "total_steps": 7496, "loss": 0.4194, "lr": 0.00010088481294913325, "epoch": 4.3810032017075775, "percentage": 54.76, "elapsed_time": "7:28:14", "remaining_time": "6:10:16"} +{"current_steps": 4110, "total_steps": 7496, "loss": 0.4206, "lr": 0.00010065197132458355, "epoch": 4.386339381003202, "percentage": 54.83, "elapsed_time": "7:28:51", "remaining_time": "6:09:47"} +{"current_steps": 4115, "total_steps": 7496, "loss": 0.4213, "lr": 0.00010041912616514814, "epoch": 4.391675560298826, "percentage": 54.9, "elapsed_time": "7:29:24", "remaining_time": "6:09:14"} +{"current_steps": 4120, "total_steps": 7496, "loss": 0.4214, "lr": 0.00010018627873327677, "epoch": 4.397011739594451, "percentage": 54.96, "elapsed_time": "7:29:58", "remaining_time": "6:08:43"} +{"current_steps": 4125, "total_steps": 7496, "loss": 0.4204, "lr": 9.995343029143144e-05, "epoch": 4.402347918890075, "percentage": 55.03, "elapsed_time": "7:30:33", "remaining_time": "6:08:11"} +{"current_steps": 4130, "total_steps": 7496, "loss": 0.42, "lr": 9.972058210207977e-05, "epoch": 4.407684098185699, "percentage": 55.1, "elapsed_time": "7:31:10", "remaining_time": "6:07:42"} +{"current_steps": 4135, "total_steps": 7496, "loss": 0.4184, "lr": 9.948773542768784e-05, "epoch": 4.413020277481324, "percentage": 55.16, "elapsed_time": "7:31:44", "remaining_time": "6:07:11"} +{"current_steps": 4140, "total_steps": 7496, "loss": 0.4221, "lr": 9.925489153071368e-05, "epoch": 4.418356456776948, "percentage": 55.23, "elapsed_time": "7:32:49", "remaining_time": "6:07:03"} +{"current_steps": 4145, "total_steps": 7496, "loss": 0.4185, "lr": 9.902205167360015e-05, "epoch": 4.423692636072572, "percentage": 55.3, "elapsed_time": "7:33:26", "remaining_time": "6:06:35"} +{"current_steps": 4150, "total_steps": 7496, "loss": 0.4193, "lr": 9.878921711876828e-05, "epoch": 4.429028815368197, "percentage": 55.36, "elapsed_time": "7:34:03", "remaining_time": "6:06:05"} +{"current_steps": 4155, "total_steps": 7496, "loss": 0.4213, "lr": 9.855638912861026e-05, "epoch": 4.434364994663821, "percentage": 55.43, "elapsed_time": "7:34:38", "remaining_time": "6:05:34"} +{"current_steps": 4160, "total_steps": 7496, "loss": 0.4207, "lr": 9.832356896548279e-05, "epoch": 4.439701173959445, "percentage": 55.5, "elapsed_time": "7:35:13", "remaining_time": "6:05:03"} +{"current_steps": 4165, "total_steps": 7496, "loss": 0.4221, "lr": 9.809075789170009e-05, "epoch": 4.44503735325507, "percentage": 55.56, "elapsed_time": "7:35:48", "remaining_time": "6:04:32"} +{"current_steps": 4170, "total_steps": 7496, "loss": 0.4205, "lr": 9.785795716952704e-05, "epoch": 4.450373532550694, "percentage": 55.63, "elapsed_time": "7:36:21", "remaining_time": "6:03:59"} +{"current_steps": 4175, "total_steps": 7496, "loss": 0.4204, "lr": 9.76251680611725e-05, "epoch": 4.455709711846318, "percentage": 55.7, "elapsed_time": "7:36:57", "remaining_time": "6:03:29"} +{"current_steps": 4180, "total_steps": 7496, "loss": 0.4195, "lr": 9.739239182878228e-05, "epoch": 4.461045891141943, "percentage": 55.76, "elapsed_time": "7:37:34", "remaining_time": "6:02:59"} +{"current_steps": 4185, "total_steps": 7496, "loss": 0.4217, "lr": 9.715962973443244e-05, "epoch": 4.466382070437566, "percentage": 55.83, "elapsed_time": "7:38:08", "remaining_time": "6:02:27"} +{"current_steps": 4190, "total_steps": 7496, "loss": 0.4199, "lr": 9.692688304012239e-05, "epoch": 4.471718249733191, "percentage": 55.9, "elapsed_time": "7:38:41", "remaining_time": "6:01:54"} +{"current_steps": 4195, "total_steps": 7496, "loss": 0.4193, "lr": 9.669415300776791e-05, "epoch": 4.477054429028815, "percentage": 55.96, "elapsed_time": "7:39:19", "remaining_time": "6:01:26"} +{"current_steps": 4200, "total_steps": 7496, "loss": 0.4211, "lr": 9.646144089919462e-05, "epoch": 4.482390608324439, "percentage": 56.03, "elapsed_time": "7:39:56", "remaining_time": "6:00:56"} +{"current_steps": 4205, "total_steps": 7496, "loss": 0.4186, "lr": 9.622874797613084e-05, "epoch": 4.487726787620064, "percentage": 56.1, "elapsed_time": "7:40:41", "remaining_time": "6:00:33"} +{"current_steps": 4210, "total_steps": 7496, "loss": 0.4201, "lr": 9.599607550020097e-05, "epoch": 4.493062966915688, "percentage": 56.16, "elapsed_time": "7:41:18", "remaining_time": "6:00:04"} +{"current_steps": 4215, "total_steps": 7496, "loss": 0.4208, "lr": 9.576342473291842e-05, "epoch": 4.498399146211312, "percentage": 56.23, "elapsed_time": "7:41:53", "remaining_time": "5:59:32"} +{"current_steps": 4220, "total_steps": 7496, "loss": 0.4213, "lr": 9.5530796935679e-05, "epoch": 4.503735325506937, "percentage": 56.3, "elapsed_time": "7:42:26", "remaining_time": "5:58:59"} +{"current_steps": 4225, "total_steps": 7496, "loss": 0.4192, "lr": 9.529819336975399e-05, "epoch": 4.509071504802561, "percentage": 56.36, "elapsed_time": "7:43:02", "remaining_time": "5:58:29"} +{"current_steps": 4230, "total_steps": 7496, "loss": 0.4196, "lr": 9.506561529628315e-05, "epoch": 4.5144076840981855, "percentage": 56.43, "elapsed_time": "7:43:39", "remaining_time": "5:57:59"} +{"current_steps": 4235, "total_steps": 7496, "loss": 0.4201, "lr": 9.48330639762682e-05, "epoch": 4.51974386339381, "percentage": 56.5, "elapsed_time": "7:44:12", "remaining_time": "5:57:26"} +{"current_steps": 4240, "total_steps": 7496, "loss": 0.4178, "lr": 9.46005406705657e-05, "epoch": 4.525080042689434, "percentage": 56.56, "elapsed_time": "7:44:46", "remaining_time": "5:56:54"} +{"current_steps": 4245, "total_steps": 7496, "loss": 0.4208, "lr": 9.436804663988041e-05, "epoch": 4.5304162219850586, "percentage": 56.63, "elapsed_time": "7:45:23", "remaining_time": "5:56:25"} +{"current_steps": 4250, "total_steps": 7496, "loss": 0.4213, "lr": 9.413558314475825e-05, "epoch": 4.535752401280683, "percentage": 56.7, "elapsed_time": "7:46:00", "remaining_time": "5:55:54"} +{"current_steps": 4255, "total_steps": 7496, "loss": 0.4207, "lr": 9.390315144557961e-05, "epoch": 4.541088580576307, "percentage": 56.76, "elapsed_time": "7:46:34", "remaining_time": "5:55:23"} +{"current_steps": 4260, "total_steps": 7496, "loss": 0.4187, "lr": 9.367075280255262e-05, "epoch": 4.546424759871932, "percentage": 56.83, "elapsed_time": "7:47:10", "remaining_time": "5:54:52"} +{"current_steps": 4265, "total_steps": 7496, "loss": 0.4202, "lr": 9.343838847570594e-05, "epoch": 4.551760939167556, "percentage": 56.9, "elapsed_time": "7:47:48", "remaining_time": "5:54:23"} +{"current_steps": 4270, "total_steps": 7496, "loss": 0.4192, "lr": 9.320605972488247e-05, "epoch": 4.55709711846318, "percentage": 56.96, "elapsed_time": "7:48:23", "remaining_time": "5:53:52"} +{"current_steps": 4275, "total_steps": 7496, "loss": 0.4186, "lr": 9.297376780973192e-05, "epoch": 4.562433297758805, "percentage": 57.03, "elapsed_time": "7:48:58", "remaining_time": "5:53:20"} +{"current_steps": 4280, "total_steps": 7496, "loss": 0.4191, "lr": 9.274151398970456e-05, "epoch": 4.567769477054429, "percentage": 57.1, "elapsed_time": "7:49:36", "remaining_time": "5:52:51"} +{"current_steps": 4285, "total_steps": 7496, "loss": 0.4166, "lr": 9.250929952404392e-05, "epoch": 4.573105656350053, "percentage": 57.16, "elapsed_time": "7:50:11", "remaining_time": "5:52:20"} +{"current_steps": 4290, "total_steps": 7496, "loss": 0.4204, "lr": 9.227712567178018e-05, "epoch": 4.578441835645678, "percentage": 57.23, "elapsed_time": "7:50:47", "remaining_time": "5:51:50"} +{"current_steps": 4295, "total_steps": 7496, "loss": 0.4198, "lr": 9.204499369172343e-05, "epoch": 4.583778014941302, "percentage": 57.3, "elapsed_time": "7:51:25", "remaining_time": "5:51:20"} +{"current_steps": 4300, "total_steps": 7496, "loss": 0.4197, "lr": 9.18129048424566e-05, "epoch": 4.5891141942369265, "percentage": 57.36, "elapsed_time": "7:52:01", "remaining_time": "5:50:50"} +{"current_steps": 4305, "total_steps": 7496, "loss": 0.4194, "lr": 9.158086038232894e-05, "epoch": 4.594450373532551, "percentage": 57.43, "elapsed_time": "7:52:47", "remaining_time": "5:50:27"} +{"current_steps": 4310, "total_steps": 7496, "loss": 0.4205, "lr": 9.134886156944883e-05, "epoch": 4.599786552828175, "percentage": 57.5, "elapsed_time": "7:53:23", "remaining_time": "5:49:56"} +{"current_steps": 4315, "total_steps": 7496, "loss": 0.4205, "lr": 9.111690966167727e-05, "epoch": 4.6051227321237995, "percentage": 57.56, "elapsed_time": "7:53:59", "remaining_time": "5:49:25"} +{"current_steps": 4320, "total_steps": 7496, "loss": 0.4191, "lr": 9.088500591662099e-05, "epoch": 4.610458911419424, "percentage": 57.63, "elapsed_time": "7:54:34", "remaining_time": "5:48:54"} +{"current_steps": 4325, "total_steps": 7496, "loss": 0.4203, "lr": 9.065315159162546e-05, "epoch": 4.615795090715048, "percentage": 57.7, "elapsed_time": "7:55:08", "remaining_time": "5:48:22"} +{"current_steps": 4330, "total_steps": 7496, "loss": 0.4196, "lr": 9.042134794376833e-05, "epoch": 4.621131270010673, "percentage": 57.76, "elapsed_time": "7:55:48", "remaining_time": "5:47:53"} +{"current_steps": 4335, "total_steps": 7496, "loss": 0.4166, "lr": 9.018959622985238e-05, "epoch": 4.626467449306297, "percentage": 57.83, "elapsed_time": "7:56:21", "remaining_time": "5:47:20"} +{"current_steps": 4340, "total_steps": 7496, "loss": 0.4178, "lr": 8.99578977063989e-05, "epoch": 4.631803628601921, "percentage": 57.9, "elapsed_time": "7:56:57", "remaining_time": "5:46:50"} +{"current_steps": 4345, "total_steps": 7496, "loss": 0.4186, "lr": 8.972625362964078e-05, "epoch": 4.637139807897546, "percentage": 57.96, "elapsed_time": "7:57:33", "remaining_time": "5:46:19"} +{"current_steps": 4350, "total_steps": 7496, "loss": 0.4196, "lr": 8.949466525551561e-05, "epoch": 4.64247598719317, "percentage": 58.03, "elapsed_time": "7:58:09", "remaining_time": "5:45:48"} +{"current_steps": 4355, "total_steps": 7496, "loss": 0.4194, "lr": 8.92631338396591e-05, "epoch": 4.647812166488794, "percentage": 58.1, "elapsed_time": "7:58:43", "remaining_time": "5:45:16"} +{"current_steps": 4360, "total_steps": 7496, "loss": 0.4196, "lr": 8.903166063739801e-05, "epoch": 4.653148345784419, "percentage": 58.16, "elapsed_time": "7:59:18", "remaining_time": "5:44:44"} +{"current_steps": 4365, "total_steps": 7496, "loss": 0.4206, "lr": 8.880024690374367e-05, "epoch": 4.658484525080043, "percentage": 58.23, "elapsed_time": "7:59:58", "remaining_time": "5:44:16"} +{"current_steps": 4370, "total_steps": 7496, "loss": 0.4173, "lr": 8.856889389338474e-05, "epoch": 4.6638207043756665, "percentage": 58.3, "elapsed_time": "8:00:32", "remaining_time": "5:43:44"} +{"current_steps": 4375, "total_steps": 7496, "loss": 0.4184, "lr": 8.833760286068079e-05, "epoch": 4.669156883671292, "percentage": 58.36, "elapsed_time": "8:01:08", "remaining_time": "5:43:13"} +{"current_steps": 4380, "total_steps": 7496, "loss": 0.4188, "lr": 8.81063750596554e-05, "epoch": 4.674493062966915, "percentage": 58.43, "elapsed_time": "8:01:45", "remaining_time": "5:42:44"} +{"current_steps": 4385, "total_steps": 7496, "loss": 0.4175, "lr": 8.787521174398914e-05, "epoch": 4.6798292422625405, "percentage": 58.5, "elapsed_time": "8:02:18", "remaining_time": "5:42:11"} +{"current_steps": 4390, "total_steps": 7496, "loss": 0.4186, "lr": 8.764411416701317e-05, "epoch": 4.685165421558164, "percentage": 58.56, "elapsed_time": "8:02:53", "remaining_time": "5:41:39"} +{"current_steps": 4395, "total_steps": 7496, "loss": 0.4201, "lr": 8.741308358170202e-05, "epoch": 4.690501600853788, "percentage": 58.63, "elapsed_time": "8:03:30", "remaining_time": "5:41:09"} +{"current_steps": 4400, "total_steps": 7496, "loss": 0.4184, "lr": 8.71821212406672e-05, "epoch": 4.695837780149413, "percentage": 58.7, "elapsed_time": "8:04:05", "remaining_time": "5:40:37"} +{"current_steps": 4405, "total_steps": 7496, "loss": 0.4169, "lr": 8.695122839615004e-05, "epoch": 4.701173959445037, "percentage": 58.76, "elapsed_time": "8:04:51", "remaining_time": "5:40:13"} +{"current_steps": 4410, "total_steps": 7496, "loss": 0.4178, "lr": 8.672040630001511e-05, "epoch": 4.706510138740661, "percentage": 58.83, "elapsed_time": "8:05:26", "remaining_time": "5:39:41"} +{"current_steps": 4415, "total_steps": 7496, "loss": 0.4188, "lr": 8.648965620374352e-05, "epoch": 4.711846318036286, "percentage": 58.9, "elapsed_time": "8:06:02", "remaining_time": "5:39:11"} +{"current_steps": 4420, "total_steps": 7496, "loss": 0.4176, "lr": 8.625897935842578e-05, "epoch": 4.71718249733191, "percentage": 58.96, "elapsed_time": "8:06:37", "remaining_time": "5:38:39"} +{"current_steps": 4425, "total_steps": 7496, "loss": 0.4171, "lr": 8.602837701475551e-05, "epoch": 4.722518676627534, "percentage": 59.03, "elapsed_time": "8:07:12", "remaining_time": "5:38:07"} +{"current_steps": 4430, "total_steps": 7496, "loss": 0.4168, "lr": 8.579785042302218e-05, "epoch": 4.727854855923159, "percentage": 59.1, "elapsed_time": "8:07:49", "remaining_time": "5:37:37"} +{"current_steps": 4435, "total_steps": 7496, "loss": 0.4191, "lr": 8.556740083310463e-05, "epoch": 4.733191035218783, "percentage": 59.16, "elapsed_time": "8:08:24", "remaining_time": "5:37:05"} +{"current_steps": 4440, "total_steps": 7496, "loss": 0.4176, "lr": 8.533702949446424e-05, "epoch": 4.7385272145144075, "percentage": 59.23, "elapsed_time": "8:08:58", "remaining_time": "5:36:33"} +{"current_steps": 4445, "total_steps": 7496, "loss": 0.4206, "lr": 8.510673765613803e-05, "epoch": 4.743863393810032, "percentage": 59.3, "elapsed_time": "8:09:35", "remaining_time": "5:36:02"} +{"current_steps": 4450, "total_steps": 7496, "loss": 0.4158, "lr": 8.487652656673209e-05, "epoch": 4.749199573105656, "percentage": 59.36, "elapsed_time": "8:10:11", "remaining_time": "5:35:32"} +{"current_steps": 4455, "total_steps": 7496, "loss": 0.4186, "lr": 8.464639747441459e-05, "epoch": 4.754535752401281, "percentage": 59.43, "elapsed_time": "8:10:45", "remaining_time": "5:34:59"} +{"current_steps": 4460, "total_steps": 7496, "loss": 0.4178, "lr": 8.441635162690927e-05, "epoch": 4.759871931696905, "percentage": 59.5, "elapsed_time": "8:11:19", "remaining_time": "5:34:27"} +{"current_steps": 4465, "total_steps": 7496, "loss": 0.418, "lr": 8.418639027148838e-05, "epoch": 4.765208110992529, "percentage": 59.57, "elapsed_time": "8:11:58", "remaining_time": "5:33:58"} +{"current_steps": 4470, "total_steps": 7496, "loss": 0.4169, "lr": 8.395651465496613e-05, "epoch": 4.770544290288154, "percentage": 59.63, "elapsed_time": "8:12:32", "remaining_time": "5:33:25"} +{"current_steps": 4475, "total_steps": 7496, "loss": 0.4153, "lr": 8.372672602369191e-05, "epoch": 4.775880469583778, "percentage": 59.7, "elapsed_time": "8:13:07", "remaining_time": "5:32:53"} +{"current_steps": 4480, "total_steps": 7496, "loss": 0.4179, "lr": 8.34970256235434e-05, "epoch": 4.781216648879402, "percentage": 59.77, "elapsed_time": "8:13:47", "remaining_time": "5:32:25"} +{"current_steps": 4485, "total_steps": 7496, "loss": 0.4165, "lr": 8.326741469992001e-05, "epoch": 4.786552828175027, "percentage": 59.83, "elapsed_time": "8:14:21", "remaining_time": "5:31:53"} +{"current_steps": 4490, "total_steps": 7496, "loss": 0.4175, "lr": 8.303789449773593e-05, "epoch": 4.791889007470651, "percentage": 59.9, "elapsed_time": "8:14:56", "remaining_time": "5:31:21"} +{"current_steps": 4495, "total_steps": 7496, "loss": 0.4182, "lr": 8.28084662614135e-05, "epoch": 4.797225186766275, "percentage": 59.97, "elapsed_time": "8:15:35", "remaining_time": "5:30:52"} +{"current_steps": 4500, "total_steps": 7496, "loss": 0.4169, "lr": 8.257913123487646e-05, "epoch": 4.8025613660619, "percentage": 60.03, "elapsed_time": "8:16:10", "remaining_time": "5:30:20"} +{"current_steps": 4505, "total_steps": 7496, "loss": 0.4184, "lr": 8.234989066154314e-05, "epoch": 4.807897545357524, "percentage": 60.1, "elapsed_time": "8:16:55", "remaining_time": "5:29:55"} +{"current_steps": 4510, "total_steps": 7496, "loss": 0.4152, "lr": 8.212074578431983e-05, "epoch": 4.8132337246531485, "percentage": 60.17, "elapsed_time": "8:17:32", "remaining_time": "5:29:24"} +{"current_steps": 4515, "total_steps": 7496, "loss": 0.4176, "lr": 8.189169784559388e-05, "epoch": 4.818569903948773, "percentage": 60.23, "elapsed_time": "8:18:08", "remaining_time": "5:28:53"} +{"current_steps": 4520, "total_steps": 7496, "loss": 0.416, "lr": 8.166274808722717e-05, "epoch": 4.823906083244397, "percentage": 60.3, "elapsed_time": "8:18:43", "remaining_time": "5:28:22"} +{"current_steps": 4525, "total_steps": 7496, "loss": 0.4144, "lr": 8.143389775054913e-05, "epoch": 4.8292422625400215, "percentage": 60.37, "elapsed_time": "8:19:18", "remaining_time": "5:27:50"} +{"current_steps": 4530, "total_steps": 7496, "loss": 0.4166, "lr": 8.120514807635019e-05, "epoch": 4.834578441835646, "percentage": 60.43, "elapsed_time": "8:19:56", "remaining_time": "5:27:20"} +{"current_steps": 4535, "total_steps": 7496, "loss": 0.4174, "lr": 8.097650030487509e-05, "epoch": 4.83991462113127, "percentage": 60.5, "elapsed_time": "8:20:32", "remaining_time": "5:26:49"} +{"current_steps": 4540, "total_steps": 7496, "loss": 0.4187, "lr": 8.07479556758159e-05, "epoch": 4.845250800426895, "percentage": 60.57, "elapsed_time": "8:21:06", "remaining_time": "5:26:16"} +{"current_steps": 4545, "total_steps": 7496, "loss": 0.4174, "lr": 8.051951542830565e-05, "epoch": 4.850586979722519, "percentage": 60.63, "elapsed_time": "8:21:44", "remaining_time": "5:25:46"} +{"current_steps": 4550, "total_steps": 7496, "loss": 0.417, "lr": 8.029118080091133e-05, "epoch": 4.855923159018143, "percentage": 60.7, "elapsed_time": "8:22:21", "remaining_time": "5:25:15"} +{"current_steps": 4555, "total_steps": 7496, "loss": 0.4168, "lr": 8.006295303162723e-05, "epoch": 4.861259338313768, "percentage": 60.77, "elapsed_time": "8:22:57", "remaining_time": "5:24:44"} +{"current_steps": 4560, "total_steps": 7496, "loss": 0.4174, "lr": 7.983483335786838e-05, "epoch": 4.866595517609392, "percentage": 60.83, "elapsed_time": "8:23:32", "remaining_time": "5:24:12"} +{"current_steps": 4565, "total_steps": 7496, "loss": 0.4174, "lr": 7.960682301646367e-05, "epoch": 4.871931696905016, "percentage": 60.9, "elapsed_time": "8:24:13", "remaining_time": "5:23:44"} +{"current_steps": 4570, "total_steps": 7496, "loss": 0.4167, "lr": 7.937892324364927e-05, "epoch": 4.877267876200641, "percentage": 60.97, "elapsed_time": "8:24:48", "remaining_time": "5:23:12"} +{"current_steps": 4575, "total_steps": 7496, "loss": 0.4191, "lr": 7.915113527506172e-05, "epoch": 4.882604055496264, "percentage": 61.03, "elapsed_time": "8:25:24", "remaining_time": "5:22:41"} +{"current_steps": 4580, "total_steps": 7496, "loss": 0.4146, "lr": 7.892346034573155e-05, "epoch": 4.887940234791889, "percentage": 61.1, "elapsed_time": "8:26:02", "remaining_time": "5:22:10"} +{"current_steps": 4585, "total_steps": 7496, "loss": 0.419, "lr": 7.869589969007631e-05, "epoch": 4.893276414087513, "percentage": 61.17, "elapsed_time": "8:26:40", "remaining_time": "5:21:40"} +{"current_steps": 4590, "total_steps": 7496, "loss": 0.4169, "lr": 7.846845454189395e-05, "epoch": 4.898612593383138, "percentage": 61.23, "elapsed_time": "8:27:13", "remaining_time": "5:21:07"} +{"current_steps": 4595, "total_steps": 7496, "loss": 0.4167, "lr": 7.82411261343563e-05, "epoch": 4.903948772678762, "percentage": 61.3, "elapsed_time": "8:27:50", "remaining_time": "5:20:37"} +{"current_steps": 4600, "total_steps": 7496, "loss": 0.4156, "lr": 7.801391570000203e-05, "epoch": 4.909284951974386, "percentage": 61.37, "elapsed_time": "8:28:25", "remaining_time": "5:20:05"} +{"current_steps": 4605, "total_steps": 7496, "loss": 0.4165, "lr": 7.778682447073037e-05, "epoch": 4.91462113127001, "percentage": 61.43, "elapsed_time": "8:29:13", "remaining_time": "5:19:41"} +{"current_steps": 4610, "total_steps": 7496, "loss": 0.416, "lr": 7.755985367779417e-05, "epoch": 4.919957310565635, "percentage": 61.5, "elapsed_time": "8:29:46", "remaining_time": "5:19:08"} +{"current_steps": 4615, "total_steps": 7496, "loss": 0.418, "lr": 7.73330045517932e-05, "epoch": 4.925293489861259, "percentage": 61.57, "elapsed_time": "8:30:23", "remaining_time": "5:18:37"} +{"current_steps": 4620, "total_steps": 7496, "loss": 0.4181, "lr": 7.710627832266772e-05, "epoch": 4.930629669156883, "percentage": 61.63, "elapsed_time": "8:30:59", "remaining_time": "5:18:05"} +{"current_steps": 4625, "total_steps": 7496, "loss": 0.4154, "lr": 7.687967621969156e-05, "epoch": 4.935965848452508, "percentage": 61.7, "elapsed_time": "8:31:34", "remaining_time": "5:17:33"} +{"current_steps": 4630, "total_steps": 7496, "loss": 0.419, "lr": 7.665319947146567e-05, "epoch": 4.941302027748132, "percentage": 61.77, "elapsed_time": "8:32:12", "remaining_time": "5:17:03"} +{"current_steps": 4635, "total_steps": 7496, "loss": 0.4147, "lr": 7.642684930591118e-05, "epoch": 4.9466382070437565, "percentage": 61.83, "elapsed_time": "8:32:47", "remaining_time": "5:16:31"} +{"current_steps": 4640, "total_steps": 7496, "loss": 0.4144, "lr": 7.620062695026305e-05, "epoch": 4.951974386339381, "percentage": 61.9, "elapsed_time": "8:33:21", "remaining_time": "5:15:58"} +{"current_steps": 4645, "total_steps": 7496, "loss": 0.4177, "lr": 7.597453363106323e-05, "epoch": 4.957310565635005, "percentage": 61.97, "elapsed_time": "8:33:59", "remaining_time": "5:15:28"} +{"current_steps": 4650, "total_steps": 7496, "loss": 0.4144, "lr": 7.574857057415401e-05, "epoch": 4.9626467449306295, "percentage": 62.03, "elapsed_time": "8:34:33", "remaining_time": "5:14:56"} +{"current_steps": 4655, "total_steps": 7496, "loss": 0.4154, "lr": 7.552273900467149e-05, "epoch": 4.967982924226254, "percentage": 62.1, "elapsed_time": "8:35:09", "remaining_time": "5:14:24"} +{"current_steps": 4660, "total_steps": 7496, "loss": 0.4175, "lr": 7.529704014703878e-05, "epoch": 4.973319103521878, "percentage": 62.17, "elapsed_time": "8:35:45", "remaining_time": "5:13:52"} +{"current_steps": 4665, "total_steps": 7496, "loss": 0.4155, "lr": 7.50714752249596e-05, "epoch": 4.978655282817503, "percentage": 62.23, "elapsed_time": "8:36:23", "remaining_time": "5:13:22"} +{"current_steps": 4670, "total_steps": 7496, "loss": 0.4163, "lr": 7.484604546141132e-05, "epoch": 4.983991462113127, "percentage": 62.3, "elapsed_time": "8:36:57", "remaining_time": "5:12:50"} +{"current_steps": 4675, "total_steps": 7496, "loss": 0.4154, "lr": 7.462075207863856e-05, "epoch": 4.989327641408751, "percentage": 62.37, "elapsed_time": "8:37:33", "remaining_time": "5:12:18"} +{"current_steps": 4680, "total_steps": 7496, "loss": 0.4143, "lr": 7.439559629814657e-05, "epoch": 4.994663820704376, "percentage": 62.43, "elapsed_time": "8:38:09", "remaining_time": "5:11:47"} +{"current_steps": 4685, "total_steps": 7496, "loss": 0.4132, "lr": 7.417057934069445e-05, "epoch": 5.0, "percentage": 62.5, "elapsed_time": "8:38:45", "remaining_time": "5:11:15"} +{"current_steps": 4690, "total_steps": 7496, "loss": 0.4111, "lr": 7.394570242628875e-05, "epoch": 5.005336179295624, "percentage": 62.57, "elapsed_time": "8:39:21", "remaining_time": "5:10:43"} +{"current_steps": 4695, "total_steps": 7496, "loss": 0.4106, "lr": 7.372096677417652e-05, "epoch": 5.010672358591249, "percentage": 62.63, "elapsed_time": "8:40:22", "remaining_time": "5:10:26"} +{"current_steps": 4700, "total_steps": 7496, "loss": 0.4112, "lr": 7.349637360283912e-05, "epoch": 5.016008537886873, "percentage": 62.7, "elapsed_time": "8:40:58", "remaining_time": "5:09:55"} +{"current_steps": 4705, "total_steps": 7496, "loss": 0.4119, "lr": 7.327192412998531e-05, "epoch": 5.021344717182497, "percentage": 62.77, "elapsed_time": "8:42:16", "remaining_time": "5:09:48"} +{"current_steps": 4710, "total_steps": 7496, "loss": 0.4126, "lr": 7.304761957254468e-05, "epoch": 5.026680896478122, "percentage": 62.83, "elapsed_time": "8:42:55", "remaining_time": "5:09:18"} +{"current_steps": 4715, "total_steps": 7496, "loss": 0.4116, "lr": 7.282346114666125e-05, "epoch": 5.032017075773746, "percentage": 62.9, "elapsed_time": "8:43:32", "remaining_time": "5:08:47"} +{"current_steps": 4720, "total_steps": 7496, "loss": 0.4102, "lr": 7.259945006768661e-05, "epoch": 5.0373532550693705, "percentage": 62.97, "elapsed_time": "8:44:06", "remaining_time": "5:08:14"} +{"current_steps": 4725, "total_steps": 7496, "loss": 0.4132, "lr": 7.237558755017358e-05, "epoch": 5.042689434364995, "percentage": 63.03, "elapsed_time": "8:44:42", "remaining_time": "5:07:42"} +{"current_steps": 4730, "total_steps": 7496, "loss": 0.4111, "lr": 7.215187480786934e-05, "epoch": 5.048025613660619, "percentage": 63.1, "elapsed_time": "8:45:18", "remaining_time": "5:07:11"} +{"current_steps": 4735, "total_steps": 7496, "loss": 0.4114, "lr": 7.192831305370913e-05, "epoch": 5.0533617929562435, "percentage": 63.17, "elapsed_time": "8:45:53", "remaining_time": "5:06:39"} +{"current_steps": 4740, "total_steps": 7496, "loss": 0.4105, "lr": 7.170490349980961e-05, "epoch": 5.058697972251868, "percentage": 63.23, "elapsed_time": "8:46:28", "remaining_time": "5:06:06"} +{"current_steps": 4745, "total_steps": 7496, "loss": 0.4103, "lr": 7.148164735746206e-05, "epoch": 5.064034151547492, "percentage": 63.3, "elapsed_time": "8:47:05", "remaining_time": "5:05:35"} +{"current_steps": 4750, "total_steps": 7496, "loss": 0.4096, "lr": 7.125854583712615e-05, "epoch": 5.069370330843117, "percentage": 63.37, "elapsed_time": "8:47:42", "remaining_time": "5:05:04"} +{"current_steps": 4755, "total_steps": 7496, "loss": 0.4122, "lr": 7.103560014842308e-05, "epoch": 5.074706510138741, "percentage": 63.43, "elapsed_time": "8:48:14", "remaining_time": "5:04:30"} +{"current_steps": 4760, "total_steps": 7496, "loss": 0.4112, "lr": 7.08128115001293e-05, "epoch": 5.080042689434365, "percentage": 63.5, "elapsed_time": "8:48:51", "remaining_time": "5:03:58"} +{"current_steps": 4765, "total_steps": 7496, "loss": 0.4112, "lr": 7.059018110016973e-05, "epoch": 5.08537886872999, "percentage": 63.57, "elapsed_time": "8:49:27", "remaining_time": "5:03:27"} +{"current_steps": 4770, "total_steps": 7496, "loss": 0.4126, "lr": 7.036771015561121e-05, "epoch": 5.090715048025614, "percentage": 63.63, "elapsed_time": "8:50:36", "remaining_time": "5:03:14"} +{"current_steps": 4775, "total_steps": 7496, "loss": 0.4094, "lr": 7.01453998726562e-05, "epoch": 5.096051227321238, "percentage": 63.7, "elapsed_time": "8:51:12", "remaining_time": "5:02:42"} +{"current_steps": 4780, "total_steps": 7496, "loss": 0.4123, "lr": 6.992325145663598e-05, "epoch": 5.101387406616863, "percentage": 63.77, "elapsed_time": "8:53:01", "remaining_time": "5:02:51"} +{"current_steps": 4785, "total_steps": 7496, "loss": 0.4105, "lr": 6.970126611200427e-05, "epoch": 5.106723585912487, "percentage": 63.83, "elapsed_time": "8:53:36", "remaining_time": "5:02:19"} +{"current_steps": 4790, "total_steps": 7496, "loss": 0.4126, "lr": 6.947944504233056e-05, "epoch": 5.112059765208111, "percentage": 63.9, "elapsed_time": "8:54:13", "remaining_time": "5:01:47"} +{"current_steps": 4795, "total_steps": 7496, "loss": 0.4113, "lr": 6.925778945029372e-05, "epoch": 5.117395944503735, "percentage": 63.97, "elapsed_time": "8:54:49", "remaining_time": "5:01:15"} +{"current_steps": 4800, "total_steps": 7496, "loss": 0.4123, "lr": 6.903630053767547e-05, "epoch": 5.122732123799359, "percentage": 64.03, "elapsed_time": "8:55:48", "remaining_time": "5:00:56"} +{"current_steps": 4805, "total_steps": 7496, "loss": 0.4122, "lr": 6.881497950535372e-05, "epoch": 5.128068303094984, "percentage": 64.1, "elapsed_time": "8:56:59", "remaining_time": "5:00:44"} +{"current_steps": 4810, "total_steps": 7496, "loss": 0.4118, "lr": 6.859382755329623e-05, "epoch": 5.133404482390608, "percentage": 64.17, "elapsed_time": "8:57:36", "remaining_time": "5:00:12"} +{"current_steps": 4815, "total_steps": 7496, "loss": 0.4135, "lr": 6.837284588055401e-05, "epoch": 5.138740661686232, "percentage": 64.23, "elapsed_time": "8:58:10", "remaining_time": "4:59:39"} +{"current_steps": 4820, "total_steps": 7496, "loss": 0.4103, "lr": 6.81520356852549e-05, "epoch": 5.144076840981857, "percentage": 64.3, "elapsed_time": "8:58:45", "remaining_time": "4:59:06"} +{"current_steps": 4825, "total_steps": 7496, "loss": 0.4116, "lr": 6.793139816459688e-05, "epoch": 5.149413020277481, "percentage": 64.37, "elapsed_time": "8:59:21", "remaining_time": "4:58:34"} +{"current_steps": 4830, "total_steps": 7496, "loss": 0.411, "lr": 6.771093451484184e-05, "epoch": 5.154749199573105, "percentage": 64.43, "elapsed_time": "8:59:55", "remaining_time": "4:58:00"} +{"current_steps": 4835, "total_steps": 7496, "loss": 0.4116, "lr": 6.749064593130896e-05, "epoch": 5.16008537886873, "percentage": 64.5, "elapsed_time": "9:00:55", "remaining_time": "4:57:42"} +{"current_steps": 4840, "total_steps": 7496, "loss": 0.4125, "lr": 6.727053360836813e-05, "epoch": 5.165421558164354, "percentage": 64.57, "elapsed_time": "9:01:58", "remaining_time": "4:57:24"} +{"current_steps": 4845, "total_steps": 7496, "loss": 0.4103, "lr": 6.705059873943375e-05, "epoch": 5.1707577374599785, "percentage": 64.63, "elapsed_time": "9:02:34", "remaining_time": "4:56:52"} +{"current_steps": 4850, "total_steps": 7496, "loss": 0.4085, "lr": 6.683084251695794e-05, "epoch": 5.176093916755603, "percentage": 64.7, "elapsed_time": "9:03:11", "remaining_time": "4:56:20"} +{"current_steps": 4855, "total_steps": 7496, "loss": 0.4125, "lr": 6.66112661324243e-05, "epoch": 5.181430096051227, "percentage": 64.77, "elapsed_time": "9:04:11", "remaining_time": "4:56:01"} +{"current_steps": 4860, "total_steps": 7496, "loss": 0.4113, "lr": 6.639187077634141e-05, "epoch": 5.1867662753468515, "percentage": 64.83, "elapsed_time": "9:07:03", "remaining_time": "4:56:42"} +{"current_steps": 4865, "total_steps": 7496, "loss": 0.4118, "lr": 6.617265763823623e-05, "epoch": 5.192102454642476, "percentage": 64.9, "elapsed_time": "9:07:40", "remaining_time": "4:56:10"} +{"current_steps": 4870, "total_steps": 7496, "loss": 0.4115, "lr": 6.59536279066479e-05, "epoch": 5.1974386339381, "percentage": 64.97, "elapsed_time": "9:08:16", "remaining_time": "4:55:38"} +{"current_steps": 4875, "total_steps": 7496, "loss": 0.4098, "lr": 6.573478276912102e-05, "epoch": 5.202774813233725, "percentage": 65.03, "elapsed_time": "9:08:51", "remaining_time": "4:55:05"} +{"current_steps": 4880, "total_steps": 7496, "loss": 0.4104, "lr": 6.551612341219951e-05, "epoch": 5.208110992529349, "percentage": 65.1, "elapsed_time": "9:09:28", "remaining_time": "4:54:33"} +{"current_steps": 4885, "total_steps": 7496, "loss": 0.4085, "lr": 6.529765102141987e-05, "epoch": 5.213447171824973, "percentage": 65.17, "elapsed_time": "9:11:52", "remaining_time": "4:54:58"} +{"current_steps": 4890, "total_steps": 7496, "loss": 0.4107, "lr": 6.507936678130493e-05, "epoch": 5.218783351120598, "percentage": 65.23, "elapsed_time": "9:12:51", "remaining_time": "4:54:38"} +{"current_steps": 4895, "total_steps": 7496, "loss": 0.4107, "lr": 6.486127187535749e-05, "epoch": 5.224119530416222, "percentage": 65.3, "elapsed_time": "9:14:20", "remaining_time": "4:54:33"} +{"current_steps": 4900, "total_steps": 7496, "loss": 0.4121, "lr": 6.464336748605368e-05, "epoch": 5.229455709711846, "percentage": 65.37, "elapsed_time": "9:16:07", "remaining_time": "4:54:37"} +{"current_steps": 4905, "total_steps": 7496, "loss": 0.4103, "lr": 6.44256547948368e-05, "epoch": 5.234791889007471, "percentage": 65.43, "elapsed_time": "9:17:28", "remaining_time": "4:54:28"} +{"current_steps": 4910, "total_steps": 7496, "loss": 0.4094, "lr": 6.420813498211071e-05, "epoch": 5.240128068303095, "percentage": 65.5, "elapsed_time": "9:18:01", "remaining_time": "4:53:53"} +{"current_steps": 4915, "total_steps": 7496, "loss": 0.4098, "lr": 6.399080922723352e-05, "epoch": 5.245464247598719, "percentage": 65.57, "elapsed_time": "9:19:30", "remaining_time": "4:53:48"} +{"current_steps": 4920, "total_steps": 7496, "loss": 0.4112, "lr": 6.377367870851124e-05, "epoch": 5.250800426894344, "percentage": 65.64, "elapsed_time": "9:20:32", "remaining_time": "4:53:29"} +{"current_steps": 4925, "total_steps": 7496, "loss": 0.4091, "lr": 6.355674460319128e-05, "epoch": 5.256136606189968, "percentage": 65.7, "elapsed_time": "9:21:07", "remaining_time": "4:52:55"} +{"current_steps": 4930, "total_steps": 7496, "loss": 0.41, "lr": 6.334000808745617e-05, "epoch": 5.2614727854855925, "percentage": 65.77, "elapsed_time": "9:22:23", "remaining_time": "4:52:42"} +{"current_steps": 4935, "total_steps": 7496, "loss": 0.409, "lr": 6.312347033641712e-05, "epoch": 5.266808964781217, "percentage": 65.84, "elapsed_time": "9:23:20", "remaining_time": "4:52:20"} +{"current_steps": 4940, "total_steps": 7496, "loss": 0.4097, "lr": 6.290713252410772e-05, "epoch": 5.272145144076841, "percentage": 65.9, "elapsed_time": "9:24:48", "remaining_time": "4:52:14"} +{"current_steps": 4945, "total_steps": 7496, "loss": 0.4096, "lr": 6.269099582347743e-05, "epoch": 5.277481323372466, "percentage": 65.97, "elapsed_time": "9:25:49", "remaining_time": "4:51:53"} +{"current_steps": 4950, "total_steps": 7496, "loss": 0.4104, "lr": 6.247506140638535e-05, "epoch": 5.28281750266809, "percentage": 66.04, "elapsed_time": "9:26:27", "remaining_time": "4:51:21"} +{"current_steps": 4955, "total_steps": 7496, "loss": 0.4117, "lr": 6.225933044359389e-05, "epoch": 5.288153681963714, "percentage": 66.1, "elapsed_time": "9:28:12", "remaining_time": "4:51:22"} +{"current_steps": 4960, "total_steps": 7496, "loss": 0.4103, "lr": 6.204380410476227e-05, "epoch": 5.293489861259339, "percentage": 66.17, "elapsed_time": "9:29:36", "remaining_time": "4:51:14"} +{"current_steps": 4965, "total_steps": 7496, "loss": 0.4094, "lr": 6.182848355844035e-05, "epoch": 5.298826040554963, "percentage": 66.24, "elapsed_time": "9:31:13", "remaining_time": "4:51:11"} +{"current_steps": 4970, "total_steps": 7496, "loss": 0.4109, "lr": 6.161336997206222e-05, "epoch": 5.304162219850587, "percentage": 66.3, "elapsed_time": "9:32:29", "remaining_time": "4:50:57"} +{"current_steps": 4975, "total_steps": 7496, "loss": 0.4096, "lr": 6.13984645119397e-05, "epoch": 5.309498399146212, "percentage": 66.37, "elapsed_time": "9:34:25", "remaining_time": "4:51:04"} +{"current_steps": 4980, "total_steps": 7496, "loss": 0.4112, "lr": 6.118376834325645e-05, "epoch": 5.314834578441836, "percentage": 66.44, "elapsed_time": "9:36:12", "remaining_time": "4:51:06"} +{"current_steps": 4985, "total_steps": 7496, "loss": 0.4104, "lr": 6.0969282630061165e-05, "epoch": 5.3201707577374595, "percentage": 66.5, "elapsed_time": "9:38:07", "remaining_time": "4:51:12"} +{"current_steps": 4990, "total_steps": 7496, "loss": 0.4125, "lr": 6.0755008535261626e-05, "epoch": 5.325506937033085, "percentage": 66.57, "elapsed_time": "9:39:58", "remaining_time": "4:51:15"} +{"current_steps": 4995, "total_steps": 7496, "loss": 0.411, "lr": 6.054094722061812e-05, "epoch": 5.330843116328708, "percentage": 66.64, "elapsed_time": "9:40:34", "remaining_time": "4:50:41"} +{"current_steps": 5000, "total_steps": 7496, "loss": 0.409, "lr": 6.032709984673739e-05, "epoch": 5.336179295624333, "percentage": 66.7, "elapsed_time": "9:41:12", "remaining_time": "4:50:08"} +{"current_steps": 5005, "total_steps": 7496, "loss": 0.4104, "lr": 6.0113467573066174e-05, "epoch": 5.341515474919957, "percentage": 66.77, "elapsed_time": "9:41:55", "remaining_time": "4:49:37"} +{"current_steps": 5010, "total_steps": 7496, "loss": 0.4102, "lr": 5.990005155788495e-05, "epoch": 5.346851654215581, "percentage": 66.84, "elapsed_time": "9:42:31", "remaining_time": "4:49:03"} +{"current_steps": 5015, "total_steps": 7496, "loss": 0.4093, "lr": 5.9686852958301765e-05, "epoch": 5.352187833511206, "percentage": 66.9, "elapsed_time": "9:43:09", "remaining_time": "4:48:29"} +{"current_steps": 5020, "total_steps": 7496, "loss": 0.4124, "lr": 5.9473872930245745e-05, "epoch": 5.35752401280683, "percentage": 66.97, "elapsed_time": "9:43:44", "remaining_time": "4:47:54"} +{"current_steps": 5025, "total_steps": 7496, "loss": 0.4101, "lr": 5.9261112628461056e-05, "epoch": 5.362860192102454, "percentage": 67.04, "elapsed_time": "9:44:18", "remaining_time": "4:47:19"} +{"current_steps": 5005, "total_steps": 7496, "loss": 0.4068, "lr": 6.0113467573066174e-05, "epoch": 5.341515474919957, "percentage": 66.77, "elapsed_time": "0:22:13", "remaining_time": "0:11:03"} +{"current_steps": 5010, "total_steps": 7496, "loss": 0.4106, "lr": 5.990005155788495e-05, "epoch": 5.346851654215581, "percentage": 66.84, "elapsed_time": "0:22:40", "remaining_time": "0:11:14"} +{"current_steps": 5015, "total_steps": 7496, "loss": 0.4085, "lr": 5.9686852958301765e-05, "epoch": 5.352187833511206, "percentage": 66.9, "elapsed_time": "0:23:05", "remaining_time": "0:11:25"} +{"current_steps": 5020, "total_steps": 7496, "loss": 0.409, "lr": 5.9473872930245745e-05, "epoch": 5.35752401280683, "percentage": 66.97, "elapsed_time": "0:24:06", "remaining_time": "0:11:53"} +{"current_steps": 5025, "total_steps": 7496, "loss": 0.4078, "lr": 5.9261112628461056e-05, "epoch": 5.362860192102454, "percentage": 67.04, "elapsed_time": "0:24:31", "remaining_time": "0:12:03"} +{"current_steps": 5030, "total_steps": 7496, "loss": 0.4089, "lr": 5.9048573206500536e-05, "epoch": 5.368196371398079, "percentage": 67.1, "elapsed_time": "0:24:54", "remaining_time": "0:12:12"} +{"current_steps": 5035, "total_steps": 7496, "loss": 0.4076, "lr": 5.883625581671934e-05, "epoch": 5.373532550693703, "percentage": 67.17, "elapsed_time": "0:25:52", "remaining_time": "0:12:38"} +{"current_steps": 5040, "total_steps": 7496, "loss": 0.4089, "lr": 5.8624161610268956e-05, "epoch": 5.378868729989327, "percentage": 67.24, "elapsed_time": "0:26:16", "remaining_time": "0:12:48"} +{"current_steps": 5045, "total_steps": 7496, "loss": 0.4068, "lr": 5.8412291737090664e-05, "epoch": 5.384204909284952, "percentage": 67.3, "elapsed_time": "0:26:40", "remaining_time": "0:12:57"} +{"current_steps": 5050, "total_steps": 7496, "loss": 0.408, "lr": 5.820064734590959e-05, "epoch": 5.389541088580576, "percentage": 67.37, "elapsed_time": "0:27:31", "remaining_time": "0:13:20"} +{"current_steps": 5055, "total_steps": 7496, "loss": 0.4084, "lr": 5.798922958422819e-05, "epoch": 5.3948772678762005, "percentage": 67.44, "elapsed_time": "0:28:07", "remaining_time": "0:13:34"} +{"current_steps": 5060, "total_steps": 7496, "loss": 0.4092, "lr": 5.777803959832029e-05, "epoch": 5.400213447171825, "percentage": 67.5, "elapsed_time": "0:28:32", "remaining_time": "0:13:44"} +{"current_steps": 5065, "total_steps": 7496, "loss": 0.4084, "lr": 5.756707853322466e-05, "epoch": 5.405549626467449, "percentage": 67.57, "elapsed_time": "0:29:11", "remaining_time": "0:14:00"} +{"current_steps": 5070, "total_steps": 7496, "loss": 0.4087, "lr": 5.7356347532738906e-05, "epoch": 5.4108858057630735, "percentage": 67.64, "elapsed_time": "0:29:59", "remaining_time": "0:14:21"} +{"current_steps": 5075, "total_steps": 7496, "loss": 0.4081, "lr": 5.714584773941336e-05, "epoch": 5.416221985058698, "percentage": 67.7, "elapsed_time": "0:30:24", "remaining_time": "0:14:30"} +{"current_steps": 5080, "total_steps": 7496, "loss": 0.4098, "lr": 5.693558029454458e-05, "epoch": 5.421558164354322, "percentage": 67.77, "elapsed_time": "0:30:51", "remaining_time": "0:14:40"} +{"current_steps": 5085, "total_steps": 7496, "loss": 0.4084, "lr": 5.6725546338169675e-05, "epoch": 5.426894343649947, "percentage": 67.84, "elapsed_time": "0:31:47", "remaining_time": "0:15:04"} +{"current_steps": 5090, "total_steps": 7496, "loss": 0.4075, "lr": 5.651574700905955e-05, "epoch": 5.432230522945571, "percentage": 67.9, "elapsed_time": "0:32:13", "remaining_time": "0:15:14"} +{"current_steps": 5095, "total_steps": 7496, "loss": 0.407, "lr": 5.6306183444713066e-05, "epoch": 5.437566702241195, "percentage": 67.97, "elapsed_time": "0:32:40", "remaining_time": "0:15:23"} +{"current_steps": 5100, "total_steps": 7496, "loss": 0.4078, "lr": 5.609685678135089e-05, "epoch": 5.44290288153682, "percentage": 68.04, "elapsed_time": "0:33:30", "remaining_time": "0:15:44"} +{"current_steps": 5105, "total_steps": 7496, "loss": 0.4082, "lr": 5.588776815390913e-05, "epoch": 5.448239060832444, "percentage": 68.1, "elapsed_time": "0:34:20", "remaining_time": "0:16:04"} +{"current_steps": 5110, "total_steps": 7496, "loss": 0.4088, "lr": 5.5678918696033355e-05, "epoch": 5.453575240128068, "percentage": 68.17, "elapsed_time": "0:34:43", "remaining_time": "0:16:12"} +{"current_steps": 5115, "total_steps": 7496, "loss": 0.406, "lr": 5.5470309540072443e-05, "epoch": 5.458911419423693, "percentage": 68.24, "elapsed_time": "0:35:29", "remaining_time": "0:16:31"} +{"current_steps": 5120, "total_steps": 7496, "loss": 0.408, "lr": 5.526194181707236e-05, "epoch": 5.464247598719317, "percentage": 68.3, "elapsed_time": "0:36:14", "remaining_time": "0:16:49"} +{"current_steps": 5125, "total_steps": 7496, "loss": 0.4092, "lr": 5.505381665677001e-05, "epoch": 5.4695837780149414, "percentage": 68.37, "elapsed_time": "0:36:39", "remaining_time": "0:16:57"} +{"current_steps": 5130, "total_steps": 7496, "loss": 0.41, "lr": 5.484593518758716e-05, "epoch": 5.474919957310566, "percentage": 68.44, "elapsed_time": "0:37:16", "remaining_time": "0:17:11"} +{"current_steps": 5135, "total_steps": 7496, "loss": 0.409, "lr": 5.4638298536624445e-05, "epoch": 5.48025613660619, "percentage": 68.5, "elapsed_time": "0:38:06", "remaining_time": "0:17:31"} +{"current_steps": 5140, "total_steps": 7496, "loss": 0.4092, "lr": 5.443090782965496e-05, "epoch": 5.4855923159018145, "percentage": 68.57, "elapsed_time": "0:38:37", "remaining_time": "0:17:42"} +{"current_steps": 5145, "total_steps": 7496, "loss": 0.4079, "lr": 5.422376419111848e-05, "epoch": 5.490928495197439, "percentage": 68.64, "elapsed_time": "0:39:06", "remaining_time": "0:17:52"} +{"current_steps": 5150, "total_steps": 7496, "loss": 0.4053, "lr": 5.401686874411516e-05, "epoch": 5.496264674493063, "percentage": 68.7, "elapsed_time": "0:39:53", "remaining_time": "0:18:10"} +{"current_steps": 5155, "total_steps": 7496, "loss": 0.4089, "lr": 5.381022261039944e-05, "epoch": 5.501600853788688, "percentage": 68.77, "elapsed_time": "0:40:29", "remaining_time": "0:18:23"} +{"current_steps": 5160, "total_steps": 7496, "loss": 0.407, "lr": 5.3603826910374153e-05, "epoch": 5.506937033084312, "percentage": 68.84, "elapsed_time": "0:40:56", "remaining_time": "0:18:32"} +{"current_steps": 5165, "total_steps": 7496, "loss": 0.4085, "lr": 5.3397682763084145e-05, "epoch": 5.512273212379936, "percentage": 68.9, "elapsed_time": "0:41:40", "remaining_time": "0:18:48"} +{"current_steps": 5170, "total_steps": 7496, "loss": 0.408, "lr": 5.319179128621053e-05, "epoch": 5.517609391675561, "percentage": 68.97, "elapsed_time": "0:42:20", "remaining_time": "0:19:03"} +{"current_steps": 5175, "total_steps": 7496, "loss": 0.4082, "lr": 5.2986153596064446e-05, "epoch": 5.522945570971185, "percentage": 69.04, "elapsed_time": "0:42:48", "remaining_time": "0:19:12"} +{"current_steps": 5180, "total_steps": 7496, "loss": 0.4063, "lr": 5.278077080758106e-05, "epoch": 5.528281750266809, "percentage": 69.1, "elapsed_time": "0:43:25", "remaining_time": "0:19:24"} +{"current_steps": 5185, "total_steps": 7496, "loss": 0.4095, "lr": 5.257564403431341e-05, "epoch": 5.533617929562434, "percentage": 69.17, "elapsed_time": "0:44:08", "remaining_time": "0:19:40"} +{"current_steps": 5190, "total_steps": 7496, "loss": 0.4076, "lr": 5.23707743884265e-05, "epoch": 5.538954108858057, "percentage": 69.24, "elapsed_time": "0:44:41", "remaining_time": "0:19:51"} +{"current_steps": 5195, "total_steps": 7496, "loss": 0.4059, "lr": 5.216616298069132e-05, "epoch": 5.544290288153682, "percentage": 69.3, "elapsed_time": "0:45:14", "remaining_time": "0:20:02"} +{"current_steps": 5200, "total_steps": 7496, "loss": 0.4065, "lr": 5.196181092047858e-05, "epoch": 5.549626467449306, "percentage": 69.37, "elapsed_time": "0:45:59", "remaining_time": "0:20:18"} +{"current_steps": 5205, "total_steps": 7496, "loss": 0.4092, "lr": 5.1757719315753e-05, "epoch": 5.554962646744931, "percentage": 69.44, "elapsed_time": "0:46:54", "remaining_time": "0:20:38"} +{"current_steps": 5210, "total_steps": 7496, "loss": 0.4081, "lr": 5.155388927306707e-05, "epoch": 5.560298826040555, "percentage": 69.5, "elapsed_time": "0:47:24", "remaining_time": "0:20:48"} +{"current_steps": 5215, "total_steps": 7496, "loss": 0.4072, "lr": 5.13503218975551e-05, "epoch": 5.565635005336179, "percentage": 69.57, "elapsed_time": "0:48:09", "remaining_time": "0:21:03"} +{"current_steps": 5220, "total_steps": 7496, "loss": 0.4071, "lr": 5.1147018292927384e-05, "epoch": 5.570971184631803, "percentage": 69.64, "elapsed_time": "0:48:52", "remaining_time": "0:21:18"} +{"current_steps": 5225, "total_steps": 7496, "loss": 0.4079, "lr": 5.094397956146391e-05, "epoch": 5.576307363927428, "percentage": 69.7, "elapsed_time": "0:49:21", "remaining_time": "0:21:27"} +{"current_steps": 5230, "total_steps": 7496, "loss": 0.4058, "lr": 5.074120680400877e-05, "epoch": 5.581643543223052, "percentage": 69.77, "elapsed_time": "0:50:01", "remaining_time": "0:21:40"} +{"current_steps": 5235, "total_steps": 7496, "loss": 0.4092, "lr": 5.053870111996379e-05, "epoch": 5.586979722518676, "percentage": 69.84, "elapsed_time": "0:50:45", "remaining_time": "0:21:55"} +{"current_steps": 5240, "total_steps": 7496, "loss": 0.4086, "lr": 5.0336463607282915e-05, "epoch": 5.592315901814301, "percentage": 69.9, "elapsed_time": "0:51:18", "remaining_time": "0:22:05"} +{"current_steps": 5245, "total_steps": 7496, "loss": 0.4071, "lr": 5.013449536246607e-05, "epoch": 5.597652081109925, "percentage": 69.97, "elapsed_time": "0:51:54", "remaining_time": "0:22:16"} +{"current_steps": 5250, "total_steps": 7496, "loss": 0.4054, "lr": 4.9932797480553174e-05, "epoch": 5.602988260405549, "percentage": 70.04, "elapsed_time": "0:52:38", "remaining_time": "0:22:31"} +{"current_steps": 5255, "total_steps": 7496, "loss": 0.4084, "lr": 4.973137105511842e-05, "epoch": 5.608324439701174, "percentage": 70.1, "elapsed_time": "0:53:16", "remaining_time": "0:22:43"} +{"current_steps": 5260, "total_steps": 7496, "loss": 0.4078, "lr": 4.9530217178264035e-05, "epoch": 5.613660618996798, "percentage": 70.17, "elapsed_time": "0:53:48", "remaining_time": "0:22:52"} +{"current_steps": 5265, "total_steps": 7496, "loss": 0.4081, "lr": 4.932933694061469e-05, "epoch": 5.6189967982924225, "percentage": 70.24, "elapsed_time": "0:54:32", "remaining_time": "0:23:06"} +{"current_steps": 5270, "total_steps": 7496, "loss": 0.4079, "lr": 4.912873143131137e-05, "epoch": 5.624332977588047, "percentage": 70.3, "elapsed_time": "0:55:08", "remaining_time": "0:23:17"} +{"current_steps": 5275, "total_steps": 7496, "loss": 0.4061, "lr": 4.892840173800545e-05, "epoch": 5.629669156883671, "percentage": 70.37, "elapsed_time": "0:55:38", "remaining_time": "0:23:25"} +{"current_steps": 5280, "total_steps": 7496, "loss": 0.4079, "lr": 4.872834894685303e-05, "epoch": 5.6350053361792956, "percentage": 70.44, "elapsed_time": "0:56:17", "remaining_time": "0:23:37"} +{"current_steps": 5285, "total_steps": 7496, "loss": 0.4072, "lr": 4.852857414250871e-05, "epoch": 5.64034151547492, "percentage": 70.5, "elapsed_time": "0:56:58", "remaining_time": "0:23:50"} +{"current_steps": 5290, "total_steps": 7496, "loss": 0.4094, "lr": 4.832907840812007e-05, "epoch": 5.645677694770544, "percentage": 70.57, "elapsed_time": "0:57:31", "remaining_time": "0:23:59"} +{"current_steps": 5295, "total_steps": 7496, "loss": 0.4074, "lr": 4.812986282532144e-05, "epoch": 5.651013874066169, "percentage": 70.64, "elapsed_time": "0:58:07", "remaining_time": "0:24:09"} +{"current_steps": 5300, "total_steps": 7496, "loss": 0.4072, "lr": 4.7930928474228345e-05, "epoch": 5.656350053361793, "percentage": 70.7, "elapsed_time": "0:58:47", "remaining_time": "0:24:21"} +{"current_steps": 5305, "total_steps": 7496, "loss": 0.4079, "lr": 4.7732276433431514e-05, "epoch": 5.661686232657417, "percentage": 70.77, "elapsed_time": "0:59:35", "remaining_time": "0:24:36"} +{"current_steps": 5310, "total_steps": 7496, "loss": 0.4075, "lr": 4.753390777999091e-05, "epoch": 5.667022411953042, "percentage": 70.84, "elapsed_time": "1:00:11", "remaining_time": "0:24:46"} +{"current_steps": 5315, "total_steps": 7496, "loss": 0.4086, "lr": 4.733582358943016e-05, "epoch": 5.672358591248666, "percentage": 70.9, "elapsed_time": "1:00:53", "remaining_time": "0:24:59"} +{"current_steps": 5320, "total_steps": 7496, "loss": 0.407, "lr": 4.7138024935730464e-05, "epoch": 5.67769477054429, "percentage": 70.97, "elapsed_time": "1:01:32", "remaining_time": "0:25:10"} +{"current_steps": 5325, "total_steps": 7496, "loss": 0.4076, "lr": 4.6940512891324986e-05, "epoch": 5.683030949839915, "percentage": 71.04, "elapsed_time": "1:02:05", "remaining_time": "0:25:18"} +{"current_steps": 5330, "total_steps": 7496, "loss": 0.4062, "lr": 4.674328852709283e-05, "epoch": 5.688367129135539, "percentage": 71.1, "elapsed_time": "1:02:48", "remaining_time": "0:25:31"} +{"current_steps": 5335, "total_steps": 7496, "loss": 0.4064, "lr": 4.654635291235343e-05, "epoch": 5.6937033084311635, "percentage": 71.17, "elapsed_time": "1:03:28", "remaining_time": "0:25:42"} +{"current_steps": 5340, "total_steps": 7496, "loss": 0.4068, "lr": 4.634970711486068e-05, "epoch": 5.699039487726788, "percentage": 71.24, "elapsed_time": "1:04:01", "remaining_time": "0:25:51"} +{"current_steps": 5345, "total_steps": 7496, "loss": 0.4055, "lr": 4.615335220079703e-05, "epoch": 5.704375667022412, "percentage": 71.3, "elapsed_time": "1:04:40", "remaining_time": "0:26:01"} +{"current_steps": 5350, "total_steps": 7496, "loss": 0.4069, "lr": 4.595728923476792e-05, "epoch": 5.7097118463180365, "percentage": 71.37, "elapsed_time": "1:05:21", "remaining_time": "0:26:12"} +{"current_steps": 5355, "total_steps": 7496, "loss": 0.4072, "lr": 4.576151927979577e-05, "epoch": 5.715048025613661, "percentage": 71.44, "elapsed_time": "1:05:55", "remaining_time": "0:26:21"} +{"current_steps": 5360, "total_steps": 7496, "loss": 0.4078, "lr": 4.556604339731445e-05, "epoch": 5.720384204909285, "percentage": 71.5, "elapsed_time": "1:06:32", "remaining_time": "0:26:30"} +{"current_steps": 5365, "total_steps": 7496, "loss": 0.4085, "lr": 4.537086264716341e-05, "epoch": 5.72572038420491, "percentage": 71.57, "elapsed_time": "1:07:13", "remaining_time": "0:26:42"} +{"current_steps": 5370, "total_steps": 7496, "loss": 0.4054, "lr": 4.517597808758183e-05, "epoch": 5.731056563500534, "percentage": 71.64, "elapsed_time": "1:07:47", "remaining_time": "0:26:50"} +{"current_steps": 5375, "total_steps": 7496, "loss": 0.4071, "lr": 4.498139077520314e-05, "epoch": 5.736392742796158, "percentage": 71.7, "elapsed_time": "1:08:23", "remaining_time": "0:26:59"} +{"current_steps": 5380, "total_steps": 7496, "loss": 0.4061, "lr": 4.478710176504898e-05, "epoch": 5.741728922091783, "percentage": 71.77, "elapsed_time": "1:09:00", "remaining_time": "0:27:08"} +{"current_steps": 5385, "total_steps": 7496, "loss": 0.4067, "lr": 4.459311211052382e-05, "epoch": 5.747065101387407, "percentage": 71.84, "elapsed_time": "1:09:36", "remaining_time": "0:27:17"} +{"current_steps": 5390, "total_steps": 7496, "loss": 0.4058, "lr": 4.43994228634089e-05, "epoch": 5.752401280683031, "percentage": 71.91, "elapsed_time": "1:10:10", "remaining_time": "0:27:25"} +{"current_steps": 5395, "total_steps": 7496, "loss": 0.4079, "lr": 4.4206035073856834e-05, "epoch": 5.757737459978655, "percentage": 71.97, "elapsed_time": "1:10:46", "remaining_time": "0:27:33"} +{"current_steps": 5400, "total_steps": 7496, "loss": 0.4044, "lr": 4.4012949790385736e-05, "epoch": 5.76307363927428, "percentage": 72.04, "elapsed_time": "1:11:23", "remaining_time": "0:27:42"} +{"current_steps": 5405, "total_steps": 7496, "loss": 0.407, "lr": 4.382016805987351e-05, "epoch": 5.7684098185699035, "percentage": 72.11, "elapsed_time": "1:12:08", "remaining_time": "0:27:54"} +{"current_steps": 5410, "total_steps": 7496, "loss": 0.4061, "lr": 4.362769092755237e-05, "epoch": 5.773745997865529, "percentage": 72.17, "elapsed_time": "1:12:46", "remaining_time": "0:28:03"} +{"current_steps": 5415, "total_steps": 7496, "loss": 0.4065, "lr": 4.343551943700291e-05, "epoch": 5.779082177161152, "percentage": 72.24, "elapsed_time": "1:13:24", "remaining_time": "0:28:12"} +{"current_steps": 5420, "total_steps": 7496, "loss": 0.4065, "lr": 4.324365463014871e-05, "epoch": 5.784418356456777, "percentage": 72.31, "elapsed_time": "1:13:59", "remaining_time": "0:28:20"} +{"current_steps": 5425, "total_steps": 7496, "loss": 0.4068, "lr": 4.3052097547250504e-05, "epoch": 5.789754535752401, "percentage": 72.37, "elapsed_time": "1:14:35", "remaining_time": "0:28:28"} +{"current_steps": 5430, "total_steps": 7496, "loss": 0.4052, "lr": 4.2860849226900545e-05, "epoch": 5.795090715048025, "percentage": 72.44, "elapsed_time": "1:15:14", "remaining_time": "0:28:37"} +{"current_steps": 5435, "total_steps": 7496, "loss": 0.4066, "lr": 4.2669910706017124e-05, "epoch": 5.80042689434365, "percentage": 72.51, "elapsed_time": "1:15:49", "remaining_time": "0:28:45"} +{"current_steps": 5440, "total_steps": 7496, "loss": 0.4044, "lr": 4.2479283019838743e-05, "epoch": 5.805763073639274, "percentage": 72.57, "elapsed_time": "1:16:27", "remaining_time": "0:28:53"} +{"current_steps": 5445, "total_steps": 7496, "loss": 0.4063, "lr": 4.228896720191873e-05, "epoch": 5.811099252934898, "percentage": 72.64, "elapsed_time": "1:17:06", "remaining_time": "0:29:02"} +{"current_steps": 5450, "total_steps": 7496, "loss": 0.4047, "lr": 4.209896428411936e-05, "epoch": 5.816435432230523, "percentage": 72.71, "elapsed_time": "1:17:45", "remaining_time": "0:29:11"} +{"current_steps": 5455, "total_steps": 7496, "loss": 0.4056, "lr": 4.190927529660653e-05, "epoch": 5.821771611526147, "percentage": 72.77, "elapsed_time": "1:18:21", "remaining_time": "0:29:19"} +{"current_steps": 5460, "total_steps": 7496, "loss": 0.4053, "lr": 4.1719901267844086e-05, "epoch": 5.827107790821771, "percentage": 72.84, "elapsed_time": "1:18:54", "remaining_time": "0:29:25"} +{"current_steps": 5465, "total_steps": 7496, "loss": 0.4038, "lr": 4.153084322458806e-05, "epoch": 5.832443970117396, "percentage": 72.91, "elapsed_time": "1:19:34", "remaining_time": "0:29:34"} +{"current_steps": 5470, "total_steps": 7496, "loss": 0.405, "lr": 4.134210219188141e-05, "epoch": 5.83778014941302, "percentage": 72.97, "elapsed_time": "1:20:08", "remaining_time": "0:29:40"} +{"current_steps": 5475, "total_steps": 7496, "loss": 0.4063, "lr": 4.115367919304822e-05, "epoch": 5.8431163287086445, "percentage": 73.04, "elapsed_time": "1:20:42", "remaining_time": "0:29:47"} +{"current_steps": 5480, "total_steps": 7496, "loss": 0.4056, "lr": 4.0965575249688336e-05, "epoch": 5.848452508004269, "percentage": 73.11, "elapsed_time": "1:21:22", "remaining_time": "0:29:56"} +{"current_steps": 5485, "total_steps": 7496, "loss": 0.4087, "lr": 4.077779138167161e-05, "epoch": 5.853788687299893, "percentage": 73.17, "elapsed_time": "1:21:56", "remaining_time": "0:30:02"} +{"current_steps": 5490, "total_steps": 7496, "loss": 0.4055, "lr": 4.059032860713261e-05, "epoch": 5.859124866595518, "percentage": 73.24, "elapsed_time": "1:22:31", "remaining_time": "0:30:09"} +{"current_steps": 5495, "total_steps": 7496, "loss": 0.4054, "lr": 4.040318794246497e-05, "epoch": 5.864461045891142, "percentage": 73.31, "elapsed_time": "1:23:09", "remaining_time": "0:30:16"} +{"current_steps": 5500, "total_steps": 7496, "loss": 0.4066, "lr": 4.0216370402315796e-05, "epoch": 5.869797225186766, "percentage": 73.37, "elapsed_time": "1:23:46", "remaining_time": "0:30:24"} +{"current_steps": 5505, "total_steps": 7496, "loss": 0.4053, "lr": 4.002987699958042e-05, "epoch": 5.875133404482391, "percentage": 73.44, "elapsed_time": "1:24:32", "remaining_time": "0:30:34"} +{"current_steps": 5510, "total_steps": 7496, "loss": 0.4033, "lr": 3.984370874539657e-05, "epoch": 5.880469583778015, "percentage": 73.51, "elapsed_time": "1:25:11", "remaining_time": "0:30:42"} +{"current_steps": 5515, "total_steps": 7496, "loss": 0.4078, "lr": 3.965786664913923e-05, "epoch": 5.885805763073639, "percentage": 73.57, "elapsed_time": "1:25:49", "remaining_time": "0:30:49"} +{"current_steps": 5520, "total_steps": 7496, "loss": 0.4055, "lr": 3.947235171841497e-05, "epoch": 5.891141942369264, "percentage": 73.64, "elapsed_time": "1:26:23", "remaining_time": "0:30:55"} +{"current_steps": 5525, "total_steps": 7496, "loss": 0.4062, "lr": 3.928716495905642e-05, "epoch": 5.896478121664888, "percentage": 73.71, "elapsed_time": "1:27:00", "remaining_time": "0:31:02"} +{"current_steps": 5530, "total_steps": 7496, "loss": 0.4043, "lr": 3.9102307375117064e-05, "epoch": 5.901814300960512, "percentage": 73.77, "elapsed_time": "1:27:38", "remaining_time": "0:31:09"} +{"current_steps": 5535, "total_steps": 7496, "loss": 0.4033, "lr": 3.891777996886551e-05, "epoch": 5.907150480256137, "percentage": 73.84, "elapsed_time": "1:28:13", "remaining_time": "0:31:15"} +{"current_steps": 5540, "total_steps": 7496, "loss": 0.405, "lr": 3.873358374078033e-05, "epoch": 5.912486659551761, "percentage": 73.91, "elapsed_time": "1:28:49", "remaining_time": "0:31:21"} +{"current_steps": 5545, "total_steps": 7496, "loss": 0.4043, "lr": 3.854971968954436e-05, "epoch": 5.9178228388473855, "percentage": 73.97, "elapsed_time": "1:29:27", "remaining_time": "0:31:28"} +{"current_steps": 5550, "total_steps": 7496, "loss": 0.4072, "lr": 3.836618881203953e-05, "epoch": 5.92315901814301, "percentage": 74.04, "elapsed_time": "1:30:01", "remaining_time": "0:31:34"} +{"current_steps": 5555, "total_steps": 7496, "loss": 0.4034, "lr": 3.8182992103341355e-05, "epoch": 5.928495197438634, "percentage": 74.11, "elapsed_time": "1:30:37", "remaining_time": "0:31:40"} +{"current_steps": 5560, "total_steps": 7496, "loss": 0.4034, "lr": 3.800013055671343e-05, "epoch": 5.9338313767342585, "percentage": 74.17, "elapsed_time": "1:31:12", "remaining_time": "0:31:45"} +{"current_steps": 5565, "total_steps": 7496, "loss": 0.4061, "lr": 3.781760516360232e-05, "epoch": 5.939167556029883, "percentage": 74.24, "elapsed_time": "1:31:49", "remaining_time": "0:31:51"} +{"current_steps": 5570, "total_steps": 7496, "loss": 0.4051, "lr": 3.7635416913631874e-05, "epoch": 5.944503735325507, "percentage": 74.31, "elapsed_time": "1:32:26", "remaining_time": "0:31:57"} +{"current_steps": 5575, "total_steps": 7496, "loss": 0.4054, "lr": 3.745356679459803e-05, "epoch": 5.949839914621132, "percentage": 74.37, "elapsed_time": "1:32:59", "remaining_time": "0:32:02"} +{"current_steps": 5580, "total_steps": 7496, "loss": 0.4059, "lr": 3.727205579246349e-05, "epoch": 5.955176093916756, "percentage": 74.44, "elapsed_time": "1:33:38", "remaining_time": "0:32:09"} +{"current_steps": 5585, "total_steps": 7496, "loss": 0.4048, "lr": 3.70908848913523e-05, "epoch": 5.96051227321238, "percentage": 74.51, "elapsed_time": "1:34:13", "remaining_time": "0:32:14"} +{"current_steps": 5590, "total_steps": 7496, "loss": 0.4061, "lr": 3.6910055073544545e-05, "epoch": 5.965848452508004, "percentage": 74.57, "elapsed_time": "1:34:48", "remaining_time": "0:32:19"} +{"current_steps": 5595, "total_steps": 7496, "loss": 0.4047, "lr": 3.672956731947093e-05, "epoch": 5.971184631803629, "percentage": 74.64, "elapsed_time": "1:35:26", "remaining_time": "0:32:25"} +{"current_steps": 5600, "total_steps": 7496, "loss": 0.4056, "lr": 3.6549422607707664e-05, "epoch": 5.9765208110992525, "percentage": 74.71, "elapsed_time": "1:36:00", "remaining_time": "0:32:30"} +{"current_steps": 5605, "total_steps": 7496, "loss": 0.4066, "lr": 3.636962191497089e-05, "epoch": 5.981856990394878, "percentage": 74.77, "elapsed_time": "1:36:49", "remaining_time": "0:32:40"} +{"current_steps": 5610, "total_steps": 7496, "loss": 0.4047, "lr": 3.6190166216111666e-05, "epoch": 5.987193169690501, "percentage": 74.84, "elapsed_time": "1:37:27", "remaining_time": "0:32:45"} +{"current_steps": 5615, "total_steps": 7496, "loss": 0.405, "lr": 3.6011056484110525e-05, "epoch": 5.9925293489861255, "percentage": 74.91, "elapsed_time": "1:38:02", "remaining_time": "0:32:50"} +{"current_steps": 5620, "total_steps": 7496, "loss": 0.4049, "lr": 3.5832293690072125e-05, "epoch": 5.99786552828175, "percentage": 74.97, "elapsed_time": "1:38:37", "remaining_time": "0:32:55"} +{"current_steps": 5625, "total_steps": 7496, "loss": 0.4039, "lr": 3.565387880322022e-05, "epoch": 6.003201707577374, "percentage": 75.04, "elapsed_time": "1:39:18", "remaining_time": "0:33:01"} +{"current_steps": 5630, "total_steps": 7496, "loss": 0.4041, "lr": 3.547581279089216e-05, "epoch": 6.008537886872999, "percentage": 75.11, "elapsed_time": "1:40:00", "remaining_time": "0:33:08"} +{"current_steps": 5635, "total_steps": 7496, "loss": 0.4041, "lr": 3.529809661853376e-05, "epoch": 6.013874066168623, "percentage": 75.17, "elapsed_time": "1:40:42", "remaining_time": "0:33:15"} +{"current_steps": 5640, "total_steps": 7496, "loss": 0.4027, "lr": 3.5120731249694114e-05, "epoch": 6.019210245464247, "percentage": 75.24, "elapsed_time": "1:41:21", "remaining_time": "0:33:21"} +{"current_steps": 5645, "total_steps": 7496, "loss": 0.4038, "lr": 3.494371764602029e-05, "epoch": 6.024546424759872, "percentage": 75.31, "elapsed_time": "1:41:57", "remaining_time": "0:33:26"} +{"current_steps": 5650, "total_steps": 7496, "loss": 0.4029, "lr": 3.476705676725218e-05, "epoch": 6.029882604055496, "percentage": 75.37, "elapsed_time": "1:42:34", "remaining_time": "0:33:30"} +{"current_steps": 5655, "total_steps": 7496, "loss": 0.403, "lr": 3.459074957121713e-05, "epoch": 6.03521878335112, "percentage": 75.44, "elapsed_time": "1:43:10", "remaining_time": "0:33:35"} +{"current_steps": 5660, "total_steps": 7496, "loss": 0.4011, "lr": 3.441479701382504e-05, "epoch": 6.040554962646745, "percentage": 75.51, "elapsed_time": "1:43:47", "remaining_time": "0:33:40"} +{"current_steps": 5665, "total_steps": 7496, "loss": 0.4018, "lr": 3.423920004906289e-05, "epoch": 6.045891141942369, "percentage": 75.57, "elapsed_time": "1:44:23", "remaining_time": "0:33:44"} +{"current_steps": 5670, "total_steps": 7496, "loss": 0.403, "lr": 3.4063959628989684e-05, "epoch": 6.0512273212379935, "percentage": 75.64, "elapsed_time": "1:45:00", "remaining_time": "0:33:48"} +{"current_steps": 5675, "total_steps": 7496, "loss": 0.4038, "lr": 3.388907670373147e-05, "epoch": 6.056563500533618, "percentage": 75.71, "elapsed_time": "1:45:36", "remaining_time": "0:33:53"} +{"current_steps": 5680, "total_steps": 7496, "loss": 0.4035, "lr": 3.3714552221475805e-05, "epoch": 6.061899679829242, "percentage": 75.77, "elapsed_time": "1:46:12", "remaining_time": "0:33:57"} +{"current_steps": 5685, "total_steps": 7496, "loss": 0.4039, "lr": 3.354038712846698e-05, "epoch": 6.0672358591248665, "percentage": 75.84, "elapsed_time": "1:46:48", "remaining_time": "0:34:01"} +{"current_steps": 5690, "total_steps": 7496, "loss": 0.4027, "lr": 3.336658236900062e-05, "epoch": 6.072572038420491, "percentage": 75.91, "elapsed_time": "1:47:26", "remaining_time": "0:34:05"} +{"current_steps": 5695, "total_steps": 7496, "loss": 0.4035, "lr": 3.3193138885418715e-05, "epoch": 6.077908217716115, "percentage": 75.97, "elapsed_time": "1:48:01", "remaining_time": "0:34:09"} +{"current_steps": 5700, "total_steps": 7496, "loss": 0.4045, "lr": 3.302005761810453e-05, "epoch": 6.08324439701174, "percentage": 76.04, "elapsed_time": "1:48:36", "remaining_time": "0:34:13"} +{"current_steps": 5705, "total_steps": 7496, "loss": 0.4016, "lr": 3.2847339505477374e-05, "epoch": 6.088580576307364, "percentage": 76.11, "elapsed_time": "1:49:27", "remaining_time": "0:34:21"} +{"current_steps": 5710, "total_steps": 7496, "loss": 0.4052, "lr": 3.267498548398766e-05, "epoch": 6.093916755602988, "percentage": 76.17, "elapsed_time": "1:50:02", "remaining_time": "0:34:25"} +{"current_steps": 5715, "total_steps": 7496, "loss": 0.4023, "lr": 3.250299648811168e-05, "epoch": 6.099252934898613, "percentage": 76.24, "elapsed_time": "1:50:39", "remaining_time": "0:34:29"} +{"current_steps": 5720, "total_steps": 7496, "loss": 0.4019, "lr": 3.233137345034669e-05, "epoch": 6.104589114194237, "percentage": 76.31, "elapsed_time": "1:51:16", "remaining_time": "0:34:32"} +{"current_steps": 5725, "total_steps": 7496, "loss": 0.403, "lr": 3.2160117301205726e-05, "epoch": 6.109925293489861, "percentage": 76.37, "elapsed_time": "1:51:52", "remaining_time": "0:34:36"} +{"current_steps": 5730, "total_steps": 7496, "loss": 0.4023, "lr": 3.198922896921259e-05, "epoch": 6.115261472785486, "percentage": 76.44, "elapsed_time": "1:52:29", "remaining_time": "0:34:40"} +{"current_steps": 5735, "total_steps": 7496, "loss": 0.4037, "lr": 3.1818709380896905e-05, "epoch": 6.12059765208111, "percentage": 76.51, "elapsed_time": "1:53:08", "remaining_time": "0:34:44"} +{"current_steps": 5740, "total_steps": 7496, "loss": 0.4041, "lr": 3.164855946078899e-05, "epoch": 6.125933831376734, "percentage": 76.57, "elapsed_time": "1:53:43", "remaining_time": "0:34:47"} +{"current_steps": 5745, "total_steps": 7496, "loss": 0.4004, "lr": 3.147878013141492e-05, "epoch": 6.131270010672359, "percentage": 76.64, "elapsed_time": "1:54:18", "remaining_time": "0:34:50"} +{"current_steps": 5750, "total_steps": 7496, "loss": 0.4037, "lr": 3.130937231329142e-05, "epoch": 6.136606189967983, "percentage": 76.71, "elapsed_time": "1:54:57", "remaining_time": "0:34:54"} +{"current_steps": 5755, "total_steps": 7496, "loss": 0.4021, "lr": 3.114033692492093e-05, "epoch": 6.1419423692636075, "percentage": 76.77, "elapsed_time": "1:55:35", "remaining_time": "0:34:57"} +{"current_steps": 5760, "total_steps": 7496, "loss": 0.4024, "lr": 3.097167488278672e-05, "epoch": 6.147278548559232, "percentage": 76.84, "elapsed_time": "1:56:09", "remaining_time": "0:35:00"} +{"current_steps": 5765, "total_steps": 7496, "loss": 0.4041, "lr": 3.080338710134778e-05, "epoch": 6.152614727854856, "percentage": 76.91, "elapsed_time": "1:56:47", "remaining_time": "0:35:04"} +{"current_steps": 5770, "total_steps": 7496, "loss": 0.4034, "lr": 3.063547449303397e-05, "epoch": 6.1579509071504805, "percentage": 76.97, "elapsed_time": "1:57:23", "remaining_time": "0:35:06"} +{"current_steps": 5775, "total_steps": 7496, "loss": 0.402, "lr": 3.04679379682409e-05, "epoch": 6.163287086446105, "percentage": 77.04, "elapsed_time": "1:58:00", "remaining_time": "0:35:09"} +{"current_steps": 5780, "total_steps": 7496, "loss": 0.4027, "lr": 3.030077843532526e-05, "epoch": 6.168623265741729, "percentage": 77.11, "elapsed_time": "1:58:37", "remaining_time": "0:35:13"} +{"current_steps": 5785, "total_steps": 7496, "loss": 0.4034, "lr": 3.013399680059965e-05, "epoch": 6.173959445037354, "percentage": 77.17, "elapsed_time": "1:59:15", "remaining_time": "0:35:16"} +{"current_steps": 5790, "total_steps": 7496, "loss": 0.4011, "lr": 2.996759396832778e-05, "epoch": 6.179295624332978, "percentage": 77.24, "elapsed_time": "1:59:51", "remaining_time": "0:35:18"} +{"current_steps": 5795, "total_steps": 7496, "loss": 0.4033, "lr": 2.980157084071957e-05, "epoch": 6.184631803628602, "percentage": 77.31, "elapsed_time": "2:00:28", "remaining_time": "0:35:21"} +{"current_steps": 5800, "total_steps": 7496, "loss": 0.4007, "lr": 2.9635928317926277e-05, "epoch": 6.189967982924227, "percentage": 77.37, "elapsed_time": "2:01:06", "remaining_time": "0:35:24"} +{"current_steps": 5805, "total_steps": 7496, "loss": 0.4054, "lr": 2.9470667298035558e-05, "epoch": 6.19530416221985, "percentage": 77.44, "elapsed_time": "2:01:53", "remaining_time": "0:35:30"} +{"current_steps": 5810, "total_steps": 7496, "loss": 0.4027, "lr": 2.930578867706657e-05, "epoch": 6.2006403415154745, "percentage": 77.51, "elapsed_time": "2:02:31", "remaining_time": "0:35:33"} +{"current_steps": 5815, "total_steps": 7496, "loss": 0.4053, "lr": 2.9141293348965183e-05, "epoch": 6.205976520811099, "percentage": 77.57, "elapsed_time": "2:03:10", "remaining_time": "0:35:36"} +{"current_steps": 5820, "total_steps": 7496, "loss": 0.403, "lr": 2.8977182205599184e-05, "epoch": 6.211312700106723, "percentage": 77.64, "elapsed_time": "2:03:47", "remaining_time": "0:35:38"} +{"current_steps": 5825, "total_steps": 7496, "loss": 0.403, "lr": 2.8813456136753213e-05, "epoch": 6.216648879402348, "percentage": 77.71, "elapsed_time": "2:04:23", "remaining_time": "0:35:40"} +{"current_steps": 5830, "total_steps": 7496, "loss": 0.4033, "lr": 2.8650116030124353e-05, "epoch": 6.221985058697972, "percentage": 77.77, "elapsed_time": "2:05:04", "remaining_time": "0:35:44"} +{"current_steps": 5835, "total_steps": 7496, "loss": 0.4012, "lr": 2.8487162771316787e-05, "epoch": 6.227321237993596, "percentage": 77.84, "elapsed_time": "2:05:40", "remaining_time": "0:35:46"} +{"current_steps": 5840, "total_steps": 7496, "loss": 0.4018, "lr": 2.832459724383748e-05, "epoch": 6.232657417289221, "percentage": 77.91, "elapsed_time": "2:06:18", "remaining_time": "0:35:48"} +{"current_steps": 5845, "total_steps": 7496, "loss": 0.4027, "lr": 2.8162420329091034e-05, "epoch": 6.237993596584845, "percentage": 77.97, "elapsed_time": "2:06:54", "remaining_time": "0:35:50"} +{"current_steps": 5850, "total_steps": 7496, "loss": 0.404, "lr": 2.80006329063751e-05, "epoch": 6.243329775880469, "percentage": 78.04, "elapsed_time": "2:07:34", "remaining_time": "0:35:53"} +{"current_steps": 5855, "total_steps": 7496, "loss": 0.4016, "lr": 2.783923585287559e-05, "epoch": 6.248665955176094, "percentage": 78.11, "elapsed_time": "2:08:08", "remaining_time": "0:35:54"} +{"current_steps": 5860, "total_steps": 7496, "loss": 0.4032, "lr": 2.76782300436619e-05, "epoch": 6.254002134471718, "percentage": 78.18, "elapsed_time": "2:08:47", "remaining_time": "0:35:57"} +{"current_steps": 5865, "total_steps": 7496, "loss": 0.4038, "lr": 2.7517616351682153e-05, "epoch": 6.259338313767342, "percentage": 78.24, "elapsed_time": "2:09:25", "remaining_time": "0:35:59"} +{"current_steps": 5870, "total_steps": 7496, "loss": 0.4044, "lr": 2.7357395647758454e-05, "epoch": 6.264674493062967, "percentage": 78.31, "elapsed_time": "2:10:01", "remaining_time": "0:36:01"} +{"current_steps": 5875, "total_steps": 7496, "loss": 0.4035, "lr": 2.719756880058214e-05, "epoch": 6.270010672358591, "percentage": 78.38, "elapsed_time": "2:10:39", "remaining_time": "0:36:02"} +{"current_steps": 5880, "total_steps": 7496, "loss": 0.4041, "lr": 2.7038136676709268e-05, "epoch": 6.2753468516542155, "percentage": 78.44, "elapsed_time": "2:11:17", "remaining_time": "0:36:04"} +{"current_steps": 5885, "total_steps": 7496, "loss": 0.4032, "lr": 2.6879100140555592e-05, "epoch": 6.28068303094984, "percentage": 78.51, "elapsed_time": "2:11:53", "remaining_time": "0:36:06"} +{"current_steps": 5890, "total_steps": 7496, "loss": 0.4041, "lr": 2.6720460054392183e-05, "epoch": 6.286019210245464, "percentage": 78.58, "elapsed_time": "2:12:29", "remaining_time": "0:36:07"} +{"current_steps": 5895, "total_steps": 7496, "loss": 0.4033, "lr": 2.656221727834056e-05, "epoch": 6.2913553895410885, "percentage": 78.64, "elapsed_time": "2:13:07", "remaining_time": "0:36:09"} +{"current_steps": 5900, "total_steps": 7496, "loss": 0.4015, "lr": 2.640437267036815e-05, "epoch": 6.296691568836713, "percentage": 78.71, "elapsed_time": "2:13:44", "remaining_time": "0:36:10"} +{"current_steps": 5905, "total_steps": 7496, "loss": 0.4017, "lr": 2.624692708628349e-05, "epoch": 6.302027748132337, "percentage": 78.78, "elapsed_time": "2:14:32", "remaining_time": "0:36:14"} +{"current_steps": 5910, "total_steps": 7496, "loss": 0.4037, "lr": 2.6089881379731684e-05, "epoch": 6.307363927427962, "percentage": 78.84, "elapsed_time": "2:15:10", "remaining_time": "0:36:16"} +{"current_steps": 5915, "total_steps": 7496, "loss": 0.4032, "lr": 2.593323640218983e-05, "epoch": 6.312700106723586, "percentage": 78.91, "elapsed_time": "2:15:49", "remaining_time": "0:36:18"} +{"current_steps": 5920, "total_steps": 7496, "loss": 0.402, "lr": 2.577699300296229e-05, "epoch": 6.31803628601921, "percentage": 78.98, "elapsed_time": "2:16:23", "remaining_time": "0:36:18"} +{"current_steps": 5925, "total_steps": 7496, "loss": 0.4013, "lr": 2.5621152029176175e-05, "epoch": 6.323372465314835, "percentage": 79.04, "elapsed_time": "2:16:59", "remaining_time": "0:36:19"} +{"current_steps": 5930, "total_steps": 7496, "loss": 0.4028, "lr": 2.5465714325776645e-05, "epoch": 6.328708644610459, "percentage": 79.11, "elapsed_time": "2:17:39", "remaining_time": "0:36:21"} +{"current_steps": 5935, "total_steps": 7496, "loss": 0.402, "lr": 2.5310680735522385e-05, "epoch": 6.334044823906083, "percentage": 79.18, "elapsed_time": "2:18:14", "remaining_time": "0:36:21"} +{"current_steps": 5940, "total_steps": 7496, "loss": 0.4038, "lr": 2.5156052098981143e-05, "epoch": 6.339381003201708, "percentage": 79.24, "elapsed_time": "2:18:51", "remaining_time": "0:36:22"} +{"current_steps": 5945, "total_steps": 7496, "loss": 0.4032, "lr": 2.5001829254524945e-05, "epoch": 6.344717182497332, "percentage": 79.31, "elapsed_time": "2:19:30", "remaining_time": "0:36:23"} +{"current_steps": 5950, "total_steps": 7496, "loss": 0.4039, "lr": 2.4848013038325756e-05, "epoch": 6.350053361792956, "percentage": 79.38, "elapsed_time": "2:20:08", "remaining_time": "0:36:24"} +{"current_steps": 5955, "total_steps": 7496, "loss": 0.4018, "lr": 2.4694604284350875e-05, "epoch": 6.355389541088581, "percentage": 79.44, "elapsed_time": "2:20:45", "remaining_time": "0:36:25"} +{"current_steps": 5960, "total_steps": 7496, "loss": 0.4022, "lr": 2.4541603824358384e-05, "epoch": 6.360725720384205, "percentage": 79.51, "elapsed_time": "2:21:22", "remaining_time": "0:36:26"} +{"current_steps": 5965, "total_steps": 7496, "loss": 0.4015, "lr": 2.4389012487892627e-05, "epoch": 6.3660618996798295, "percentage": 79.58, "elapsed_time": "2:22:01", "remaining_time": "0:36:27"} +{"current_steps": 5970, "total_steps": 7496, "loss": 0.4022, "lr": 2.4236831102279756e-05, "epoch": 6.371398078975454, "percentage": 79.64, "elapsed_time": "2:22:37", "remaining_time": "0:36:27"} +{"current_steps": 5975, "total_steps": 7496, "loss": 0.4019, "lr": 2.408506049262329e-05, "epoch": 6.376734258271078, "percentage": 79.71, "elapsed_time": "2:23:10", "remaining_time": "0:36:26"} +{"current_steps": 5980, "total_steps": 7496, "loss": 0.401, "lr": 2.3933701481799496e-05, "epoch": 6.382070437566703, "percentage": 79.78, "elapsed_time": "2:23:48", "remaining_time": "0:36:27"} +{"current_steps": 5985, "total_steps": 7496, "loss": 0.4018, "lr": 2.3782754890453106e-05, "epoch": 6.387406616862327, "percentage": 79.84, "elapsed_time": "2:24:25", "remaining_time": "0:36:27"} +{"current_steps": 5990, "total_steps": 7496, "loss": 0.405, "lr": 2.363222153699276e-05, "epoch": 6.392742796157951, "percentage": 79.91, "elapsed_time": "2:25:00", "remaining_time": "0:36:27"} +{"current_steps": 5995, "total_steps": 7496, "loss": 0.403, "lr": 2.34821022375865e-05, "epoch": 6.398078975453576, "percentage": 79.98, "elapsed_time": "2:25:37", "remaining_time": "0:36:27"} +{"current_steps": 6000, "total_steps": 7496, "loss": 0.4023, "lr": 2.333239780615756e-05, "epoch": 6.4034151547492, "percentage": 80.04, "elapsed_time": "2:26:15", "remaining_time": "0:36:28"} +{"current_steps": 6005, "total_steps": 7496, "loss": 0.403, "lr": 2.3183109054379715e-05, "epoch": 6.408751334044824, "percentage": 80.11, "elapsed_time": "2:27:05", "remaining_time": "0:36:31"} +{"current_steps": 6010, "total_steps": 7496, "loss": 0.4012, "lr": 2.3034236791673058e-05, "epoch": 6.414087513340448, "percentage": 80.18, "elapsed_time": "2:27:43", "remaining_time": "0:36:31"} +{"current_steps": 6015, "total_steps": 7496, "loss": 0.4018, "lr": 2.2885781825199526e-05, "epoch": 6.419423692636073, "percentage": 80.24, "elapsed_time": "2:28:20", "remaining_time": "0:36:31"} +{"current_steps": 6020, "total_steps": 7496, "loss": 0.4018, "lr": 2.2737744959858577e-05, "epoch": 6.4247598719316965, "percentage": 80.31, "elapsed_time": "2:28:55", "remaining_time": "0:36:30"} +{"current_steps": 6025, "total_steps": 7496, "loss": 0.4027, "lr": 2.2590126998282725e-05, "epoch": 6.430096051227321, "percentage": 80.38, "elapsed_time": "2:29:31", "remaining_time": "0:36:30"} +{"current_steps": 6030, "total_steps": 7496, "loss": 0.4034, "lr": 2.2442928740833245e-05, "epoch": 6.435432230522945, "percentage": 80.44, "elapsed_time": "2:30:08", "remaining_time": "0:36:30"} +{"current_steps": 6035, "total_steps": 7496, "loss": 0.403, "lr": 2.229615098559594e-05, "epoch": 6.44076840981857, "percentage": 80.51, "elapsed_time": "2:30:43", "remaining_time": "0:36:29"} +{"current_steps": 6040, "total_steps": 7496, "loss": 0.4016, "lr": 2.214979452837661e-05, "epoch": 6.446104589114194, "percentage": 80.58, "elapsed_time": "2:31:20", "remaining_time": "0:36:28"} +{"current_steps": 6045, "total_steps": 7496, "loss": 0.4027, "lr": 2.200386016269691e-05, "epoch": 6.451440768409818, "percentage": 80.64, "elapsed_time": "2:31:57", "remaining_time": "0:36:28"} +{"current_steps": 6050, "total_steps": 7496, "loss": 0.4017, "lr": 2.1858348679789987e-05, "epoch": 6.456776947705443, "percentage": 80.71, "elapsed_time": "2:32:36", "remaining_time": "0:36:28"} +{"current_steps": 6055, "total_steps": 7496, "loss": 0.4016, "lr": 2.171326086859612e-05, "epoch": 6.462113127001067, "percentage": 80.78, "elapsed_time": "2:33:11", "remaining_time": "0:36:27"} +{"current_steps": 6060, "total_steps": 7496, "loss": 0.4033, "lr": 2.1568597515758572e-05, "epoch": 6.467449306296691, "percentage": 80.84, "elapsed_time": "2:33:49", "remaining_time": "0:36:26"} +{"current_steps": 6065, "total_steps": 7496, "loss": 0.4026, "lr": 2.142435940561921e-05, "epoch": 6.472785485592316, "percentage": 80.91, "elapsed_time": "2:34:28", "remaining_time": "0:36:26"} +{"current_steps": 6070, "total_steps": 7496, "loss": 0.4017, "lr": 2.128054732021435e-05, "epoch": 6.47812166488794, "percentage": 80.98, "elapsed_time": "2:35:03", "remaining_time": "0:36:25"} +{"current_steps": 6075, "total_steps": 7496, "loss": 0.4032, "lr": 2.1137162039270386e-05, "epoch": 6.483457844183564, "percentage": 81.04, "elapsed_time": "2:35:40", "remaining_time": "0:36:24"} +{"current_steps": 6080, "total_steps": 7496, "loss": 0.402, "lr": 2.0994204340199806e-05, "epoch": 6.488794023479189, "percentage": 81.11, "elapsed_time": "2:36:20", "remaining_time": "0:36:24"} +{"current_steps": 6085, "total_steps": 7496, "loss": 0.4009, "lr": 2.085167499809666e-05, "epoch": 6.494130202774813, "percentage": 81.18, "elapsed_time": "2:36:55", "remaining_time": "0:36:23"} +{"current_steps": 6090, "total_steps": 7496, "loss": 0.4043, "lr": 2.0709574785732544e-05, "epoch": 6.4994663820704375, "percentage": 81.24, "elapsed_time": "2:37:30", "remaining_time": "0:36:21"} +{"current_steps": 6095, "total_steps": 7496, "loss": 0.4023, "lr": 2.0567904473552446e-05, "epoch": 6.504802561366062, "percentage": 81.31, "elapsed_time": "2:38:10", "remaining_time": "0:36:21"} +{"current_steps": 6100, "total_steps": 7496, "loss": 0.4041, "lr": 2.0426664829670415e-05, "epoch": 6.510138740661686, "percentage": 81.38, "elapsed_time": "2:38:47", "remaining_time": "0:36:20"} +{"current_steps": 6105, "total_steps": 7496, "loss": 0.3993, "lr": 2.0285856619865573e-05, "epoch": 6.5154749199573105, "percentage": 81.44, "elapsed_time": "2:39:34", "remaining_time": "0:36:21"} +{"current_steps": 6110, "total_steps": 7496, "loss": 0.4041, "lr": 2.014548060757785e-05, "epoch": 6.520811099252935, "percentage": 81.51, "elapsed_time": "2:40:12", "remaining_time": "0:36:20"} +{"current_steps": 6115, "total_steps": 7496, "loss": 0.4021, "lr": 2.0005537553903787e-05, "epoch": 6.526147278548559, "percentage": 81.58, "elapsed_time": "2:40:48", "remaining_time": "0:36:19"} +{"current_steps": 6120, "total_steps": 7496, "loss": 0.4016, "lr": 1.9866028217592612e-05, "epoch": 6.531483457844184, "percentage": 81.64, "elapsed_time": "2:41:23", "remaining_time": "0:36:17"} +{"current_steps": 6125, "total_steps": 7496, "loss": 0.4025, "lr": 1.972695335504192e-05, "epoch": 6.536819637139808, "percentage": 81.71, "elapsed_time": "2:42:00", "remaining_time": "0:36:15"} +{"current_steps": 6130, "total_steps": 7496, "loss": 0.4029, "lr": 1.9588313720293717e-05, "epoch": 6.542155816435432, "percentage": 81.78, "elapsed_time": "2:42:39", "remaining_time": "0:36:14"} +{"current_steps": 6135, "total_steps": 7496, "loss": 0.4008, "lr": 1.9450110065030202e-05, "epoch": 6.547491995731057, "percentage": 81.84, "elapsed_time": "2:43:17", "remaining_time": "0:36:13"} +{"current_steps": 6140, "total_steps": 7496, "loss": 0.4024, "lr": 1.9312343138569822e-05, "epoch": 6.552828175026681, "percentage": 81.91, "elapsed_time": "2:43:53", "remaining_time": "0:36:11"} +{"current_steps": 6145, "total_steps": 7496, "loss": 0.4007, "lr": 1.9175013687863165e-05, "epoch": 6.558164354322305, "percentage": 81.98, "elapsed_time": "2:44:30", "remaining_time": "0:36:10"} +{"current_steps": 6150, "total_steps": 7496, "loss": 0.4019, "lr": 1.9038122457488804e-05, "epoch": 6.56350053361793, "percentage": 82.04, "elapsed_time": "2:45:07", "remaining_time": "0:36:08"} +{"current_steps": 6155, "total_steps": 7496, "loss": 0.4015, "lr": 1.8901670189649457e-05, "epoch": 6.568836712913554, "percentage": 82.11, "elapsed_time": "2:45:42", "remaining_time": "0:36:06"} +{"current_steps": 6160, "total_steps": 7496, "loss": 0.4034, "lr": 1.8765657624167765e-05, "epoch": 6.574172892209178, "percentage": 82.18, "elapsed_time": "2:46:21", "remaining_time": "0:36:04"} +{"current_steps": 6165, "total_steps": 7496, "loss": 0.4015, "lr": 1.863008549848243e-05, "epoch": 6.579509071504803, "percentage": 82.24, "elapsed_time": "2:46:58", "remaining_time": "0:36:03"} +{"current_steps": 6170, "total_steps": 7496, "loss": 0.4013, "lr": 1.849495454764416e-05, "epoch": 6.584845250800427, "percentage": 82.31, "elapsed_time": "2:47:35", "remaining_time": "0:36:00"} +{"current_steps": 6175, "total_steps": 7496, "loss": 0.402, "lr": 1.8360265504311635e-05, "epoch": 6.5901814300960515, "percentage": 82.38, "elapsed_time": "2:48:11", "remaining_time": "0:35:58"} +{"current_steps": 6180, "total_steps": 7496, "loss": 0.4025, "lr": 1.822601909874765e-05, "epoch": 6.595517609391676, "percentage": 82.44, "elapsed_time": "2:48:47", "remaining_time": "0:35:56"} +{"current_steps": 6185, "total_steps": 7496, "loss": 0.4013, "lr": 1.8092216058814993e-05, "epoch": 6.6008537886873, "percentage": 82.51, "elapsed_time": "2:49:21", "remaining_time": "0:35:53"} +{"current_steps": 6190, "total_steps": 7496, "loss": 0.4007, "lr": 1.7958857109972714e-05, "epoch": 6.606189967982925, "percentage": 82.58, "elapsed_time": "2:49:57", "remaining_time": "0:35:51"} +{"current_steps": 6195, "total_steps": 7496, "loss": 0.4015, "lr": 1.7825942975271946e-05, "epoch": 6.611526147278549, "percentage": 82.64, "elapsed_time": "2:50:36", "remaining_time": "0:35:49"} +{"current_steps": 6200, "total_steps": 7496, "loss": 0.4014, "lr": 1.7693474375352193e-05, "epoch": 6.616862326574173, "percentage": 82.71, "elapsed_time": "2:51:09", "remaining_time": "0:35:46"} +{"current_steps": 6205, "total_steps": 7496, "loss": 0.4018, "lr": 1.7561452028437362e-05, "epoch": 6.622198505869797, "percentage": 82.78, "elapsed_time": "2:51:56", "remaining_time": "0:35:46"} +{"current_steps": 6210, "total_steps": 7496, "loss": 0.402, "lr": 1.7429876650331732e-05, "epoch": 6.627534685165422, "percentage": 82.84, "elapsed_time": "2:52:36", "remaining_time": "0:35:44"} +{"current_steps": 6215, "total_steps": 7496, "loss": 0.4019, "lr": 1.7298748954416334e-05, "epoch": 6.6328708644610455, "percentage": 82.91, "elapsed_time": "2:53:10", "remaining_time": "0:35:41"} +{"current_steps": 6220, "total_steps": 7496, "loss": 0.4004, "lr": 1.716806965164477e-05, "epoch": 6.638207043756671, "percentage": 82.98, "elapsed_time": "2:53:44", "remaining_time": "0:35:38"} +{"current_steps": 6225, "total_steps": 7496, "loss": 0.3983, "lr": 1.7037839450539706e-05, "epoch": 6.643543223052294, "percentage": 83.04, "elapsed_time": "2:54:23", "remaining_time": "0:35:36"} +{"current_steps": 6230, "total_steps": 7496, "loss": 0.4032, "lr": 1.6908059057188686e-05, "epoch": 6.6488794023479185, "percentage": 83.11, "elapsed_time": "2:54:59", "remaining_time": "0:35:33"} +{"current_steps": 6235, "total_steps": 7496, "loss": 0.3994, "lr": 1.677872917524057e-05, "epoch": 6.654215581643543, "percentage": 83.18, "elapsed_time": "2:55:35", "remaining_time": "0:35:30"} +{"current_steps": 6240, "total_steps": 7496, "loss": 0.4022, "lr": 1.664985050590162e-05, "epoch": 6.659551760939167, "percentage": 83.24, "elapsed_time": "2:56:11", "remaining_time": "0:35:27"} +{"current_steps": 6245, "total_steps": 7496, "loss": 0.4026, "lr": 1.6521423747931597e-05, "epoch": 6.664887940234792, "percentage": 83.31, "elapsed_time": "2:56:49", "remaining_time": "0:35:25"} +{"current_steps": 6250, "total_steps": 7496, "loss": 0.4024, "lr": 1.63934495976402e-05, "epoch": 6.670224119530416, "percentage": 83.38, "elapsed_time": "2:57:24", "remaining_time": "0:35:22"} +{"current_steps": 6255, "total_steps": 7496, "loss": 0.4036, "lr": 1.626592874888303e-05, "epoch": 6.67556029882604, "percentage": 83.44, "elapsed_time": "2:58:00", "remaining_time": "0:35:19"} +{"current_steps": 6260, "total_steps": 7496, "loss": 0.4029, "lr": 1.6138861893058067e-05, "epoch": 6.680896478121665, "percentage": 83.51, "elapsed_time": "2:58:37", "remaining_time": "0:35:16"} +{"current_steps": 6265, "total_steps": 7496, "loss": 0.4023, "lr": 1.601224971910177e-05, "epoch": 6.686232657417289, "percentage": 83.58, "elapsed_time": "2:59:11", "remaining_time": "0:35:12"} +{"current_steps": 6270, "total_steps": 7496, "loss": 0.401, "lr": 1.588609291348535e-05, "epoch": 6.691568836712913, "percentage": 83.64, "elapsed_time": "2:59:47", "remaining_time": "0:35:09"} +{"current_steps": 6275, "total_steps": 7496, "loss": 0.4013, "lr": 1.5760392160211156e-05, "epoch": 6.696905016008538, "percentage": 83.71, "elapsed_time": "3:00:24", "remaining_time": "0:35:06"} +{"current_steps": 6280, "total_steps": 7496, "loss": 0.4004, "lr": 1.563514814080881e-05, "epoch": 6.702241195304162, "percentage": 83.78, "elapsed_time": "3:01:02", "remaining_time": "0:35:03"} +{"current_steps": 6285, "total_steps": 7496, "loss": 0.4022, "lr": 1.5510361534331686e-05, "epoch": 6.707577374599786, "percentage": 83.84, "elapsed_time": "3:01:38", "remaining_time": "0:34:59"} +{"current_steps": 6290, "total_steps": 7496, "loss": 0.4006, "lr": 1.538603301735305e-05, "epoch": 6.712913553895411, "percentage": 83.91, "elapsed_time": "3:02:15", "remaining_time": "0:34:56"} +{"current_steps": 6295, "total_steps": 7496, "loss": 0.403, "lr": 1.5262163263962527e-05, "epoch": 6.718249733191035, "percentage": 83.98, "elapsed_time": "3:02:50", "remaining_time": "0:34:53"} +{"current_steps": 6300, "total_steps": 7496, "loss": 0.4021, "lr": 1.5138752945762425e-05, "epoch": 6.7235859124866595, "percentage": 84.04, "elapsed_time": "3:03:27", "remaining_time": "0:34:49"} +{"current_steps": 6305, "total_steps": 7496, "loss": 0.3988, "lr": 1.5015802731863981e-05, "epoch": 6.728922091782284, "percentage": 84.11, "elapsed_time": "3:04:12", "remaining_time": "0:34:47"} +{"current_steps": 6310, "total_steps": 7496, "loss": 0.401, "lr": 1.4893313288883915e-05, "epoch": 6.734258271077908, "percentage": 84.18, "elapsed_time": "3:04:47", "remaining_time": "0:34:43"} +{"current_steps": 6315, "total_steps": 7496, "loss": 0.4005, "lr": 1.4771285280940616e-05, "epoch": 6.7395944503735326, "percentage": 84.24, "elapsed_time": "3:05:23", "remaining_time": "0:34:40"} +{"current_steps": 6320, "total_steps": 7496, "loss": 0.3999, "lr": 1.4649719369650772e-05, "epoch": 6.744930629669157, "percentage": 84.31, "elapsed_time": "3:05:58", "remaining_time": "0:34:36"} +{"current_steps": 6325, "total_steps": 7496, "loss": 0.4019, "lr": 1.4528616214125535e-05, "epoch": 6.750266808964781, "percentage": 84.38, "elapsed_time": "3:06:33", "remaining_time": "0:34:32"} +{"current_steps": 6330, "total_steps": 7496, "loss": 0.4025, "lr": 1.4407976470967133e-05, "epoch": 6.755602988260406, "percentage": 84.45, "elapsed_time": "3:07:11", "remaining_time": "0:34:28"} +{"current_steps": 6335, "total_steps": 7496, "loss": 0.4026, "lr": 1.428780079426526e-05, "epoch": 6.76093916755603, "percentage": 84.51, "elapsed_time": "3:07:47", "remaining_time": "0:34:24"} +{"current_steps": 6340, "total_steps": 7496, "loss": 0.4018, "lr": 1.4168089835593446e-05, "epoch": 6.766275346851654, "percentage": 84.58, "elapsed_time": "3:08:24", "remaining_time": "0:34:21"} +{"current_steps": 6345, "total_steps": 7496, "loss": 0.4021, "lr": 1.4048844244005666e-05, "epoch": 6.771611526147279, "percentage": 84.65, "elapsed_time": "3:09:00", "remaining_time": "0:34:17"} +{"current_steps": 6350, "total_steps": 7496, "loss": 0.4016, "lr": 1.3930064666032693e-05, "epoch": 6.776947705442903, "percentage": 84.71, "elapsed_time": "3:09:39", "remaining_time": "0:34:13"} +{"current_steps": 6355, "total_steps": 7496, "loss": 0.4037, "lr": 1.381175174567868e-05, "epoch": 6.782283884738527, "percentage": 84.78, "elapsed_time": "3:10:13", "remaining_time": "0:34:09"} +{"current_steps": 6360, "total_steps": 7496, "loss": 0.4041, "lr": 1.3693906124417655e-05, "epoch": 6.787620064034152, "percentage": 84.85, "elapsed_time": "3:10:49", "remaining_time": "0:34:05"} +{"current_steps": 6365, "total_steps": 7496, "loss": 0.4021, "lr": 1.357652844118994e-05, "epoch": 6.792956243329776, "percentage": 84.91, "elapsed_time": "3:11:28", "remaining_time": "0:34:01"} +{"current_steps": 6370, "total_steps": 7496, "loss": 0.4025, "lr": 1.3459619332398887e-05, "epoch": 6.7982924226254005, "percentage": 84.98, "elapsed_time": "3:12:03", "remaining_time": "0:33:57"} +{"current_steps": 6375, "total_steps": 7496, "loss": 0.3985, "lr": 1.3343179431907193e-05, "epoch": 6.803628601921025, "percentage": 85.05, "elapsed_time": "3:12:38", "remaining_time": "0:33:52"} +{"current_steps": 6380, "total_steps": 7496, "loss": 0.4, "lr": 1.32272093710337e-05, "epoch": 6.808964781216649, "percentage": 85.11, "elapsed_time": "3:13:16", "remaining_time": "0:33:48"} +{"current_steps": 6385, "total_steps": 7496, "loss": 0.4039, "lr": 1.311170977854973e-05, "epoch": 6.8143009605122735, "percentage": 85.18, "elapsed_time": "3:13:53", "remaining_time": "0:33:44"} +{"current_steps": 6390, "total_steps": 7496, "loss": 0.4015, "lr": 1.2996681280675894e-05, "epoch": 6.819637139807898, "percentage": 85.25, "elapsed_time": "3:14:27", "remaining_time": "0:33:39"} +{"current_steps": 6395, "total_steps": 7496, "loss": 0.4001, "lr": 1.2882124501078597e-05, "epoch": 6.824973319103522, "percentage": 85.31, "elapsed_time": "3:15:06", "remaining_time": "0:33:35"} +{"current_steps": 6400, "total_steps": 7496, "loss": 0.4002, "lr": 1.2768040060866593e-05, "epoch": 6.830309498399147, "percentage": 85.38, "elapsed_time": "3:15:44", "remaining_time": "0:33:31"} +{"current_steps": 6405, "total_steps": 7496, "loss": 0.4012, "lr": 1.2654428578587763e-05, "epoch": 6.835645677694771, "percentage": 85.45, "elapsed_time": "3:16:30", "remaining_time": "0:33:28"} +{"current_steps": 6410, "total_steps": 7496, "loss": 0.403, "lr": 1.2541290670225647e-05, "epoch": 6.840981856990394, "percentage": 85.51, "elapsed_time": "3:17:05", "remaining_time": "0:33:23"} +{"current_steps": 6415, "total_steps": 7496, "loss": 0.4013, "lr": 1.2428626949196076e-05, "epoch": 6.84631803628602, "percentage": 85.58, "elapsed_time": "3:17:43", "remaining_time": "0:33:19"} +{"current_steps": 6420, "total_steps": 7496, "loss": 0.4001, "lr": 1.231643802634409e-05, "epoch": 6.851654215581643, "percentage": 85.65, "elapsed_time": "3:18:18", "remaining_time": "0:33:14"} +{"current_steps": 6425, "total_steps": 7496, "loss": 0.3985, "lr": 1.220472450994027e-05, "epoch": 6.856990394877268, "percentage": 85.71, "elapsed_time": "3:18:54", "remaining_time": "0:33:09"} +{"current_steps": 6430, "total_steps": 7496, "loss": 0.3989, "lr": 1.2093487005677728e-05, "epoch": 6.862326574172892, "percentage": 85.78, "elapsed_time": "3:19:32", "remaining_time": "0:33:04"} +{"current_steps": 6435, "total_steps": 7496, "loss": 0.4017, "lr": 1.1982726116668642e-05, "epoch": 6.867662753468516, "percentage": 85.85, "elapsed_time": "3:20:10", "remaining_time": "0:33:00"} +{"current_steps": 6440, "total_steps": 7496, "loss": 0.4006, "lr": 1.1872442443441113e-05, "epoch": 6.8729989327641405, "percentage": 85.91, "elapsed_time": "3:20:45", "remaining_time": "0:32:55"} +{"current_steps": 6445, "total_steps": 7496, "loss": 0.4017, "lr": 1.1762636583935793e-05, "epoch": 6.878335112059765, "percentage": 85.98, "elapsed_time": "3:21:23", "remaining_time": "0:32:50"} +{"current_steps": 6450, "total_steps": 7496, "loss": 0.4011, "lr": 1.1653309133502777e-05, "epoch": 6.883671291355389, "percentage": 86.05, "elapsed_time": "3:22:01", "remaining_time": "0:32:45"} +{"current_steps": 6455, "total_steps": 7496, "loss": 0.4028, "lr": 1.1544460684898261e-05, "epoch": 6.889007470651014, "percentage": 86.11, "elapsed_time": "3:22:38", "remaining_time": "0:32:40"} +{"current_steps": 6460, "total_steps": 7496, "loss": 0.4006, "lr": 1.1436091828281348e-05, "epoch": 6.894343649946638, "percentage": 86.18, "elapsed_time": "3:23:17", "remaining_time": "0:32:36"} +{"current_steps": 6465, "total_steps": 7496, "loss": 0.4026, "lr": 1.1328203151210915e-05, "epoch": 6.899679829242262, "percentage": 86.25, "elapsed_time": "3:23:56", "remaining_time": "0:32:31"} +{"current_steps": 6470, "total_steps": 7496, "loss": 0.3998, "lr": 1.1220795238642357e-05, "epoch": 6.905016008537887, "percentage": 86.31, "elapsed_time": "3:24:35", "remaining_time": "0:32:26"} +{"current_steps": 6475, "total_steps": 7496, "loss": 0.4003, "lr": 1.1113868672924422e-05, "epoch": 6.910352187833511, "percentage": 86.38, "elapsed_time": "3:25:10", "remaining_time": "0:32:21"} +{"current_steps": 6480, "total_steps": 7496, "loss": 0.4017, "lr": 1.1007424033796087e-05, "epoch": 6.915688367129135, "percentage": 86.45, "elapsed_time": "3:25:47", "remaining_time": "0:32:15"} +{"current_steps": 6485, "total_steps": 7496, "loss": 0.4007, "lr": 1.0901461898383424e-05, "epoch": 6.92102454642476, "percentage": 86.51, "elapsed_time": "3:26:25", "remaining_time": "0:32:10"} +{"current_steps": 6490, "total_steps": 7496, "loss": 0.4019, "lr": 1.079598284119644e-05, "epoch": 6.926360725720384, "percentage": 86.58, "elapsed_time": "3:26:59", "remaining_time": "0:32:05"} +{"current_steps": 6495, "total_steps": 7496, "loss": 0.4016, "lr": 1.06909874341259e-05, "epoch": 6.931696905016008, "percentage": 86.65, "elapsed_time": "3:27:39", "remaining_time": "0:32:00"} +{"current_steps": 6500, "total_steps": 7496, "loss": 0.4009, "lr": 1.0586476246440403e-05, "epoch": 6.937033084311633, "percentage": 86.71, "elapsed_time": "3:28:15", "remaining_time": "0:31:54"} +{"current_steps": 6505, "total_steps": 7496, "loss": 0.4003, "lr": 1.0482449844783093e-05, "epoch": 6.942369263607257, "percentage": 86.78, "elapsed_time": "3:29:03", "remaining_time": "0:31:50"} +{"current_steps": 6510, "total_steps": 7496, "loss": 0.4003, "lr": 1.0378908793168718e-05, "epoch": 6.9477054429028815, "percentage": 86.85, "elapsed_time": "3:29:38", "remaining_time": "0:31:45"} +{"current_steps": 6515, "total_steps": 7496, "loss": 0.4017, "lr": 1.0275853652980571e-05, "epoch": 6.953041622198506, "percentage": 86.91, "elapsed_time": "3:30:17", "remaining_time": "0:31:39"} +{"current_steps": 6520, "total_steps": 7496, "loss": 0.4021, "lr": 1.0173284982967335e-05, "epoch": 6.95837780149413, "percentage": 86.98, "elapsed_time": "3:30:54", "remaining_time": "0:31:34"} +{"current_steps": 6525, "total_steps": 7496, "loss": 0.4019, "lr": 1.0071203339240198e-05, "epoch": 6.963713980789755, "percentage": 87.05, "elapsed_time": "3:31:33", "remaining_time": "0:31:28"} +{"current_steps": 6530, "total_steps": 7496, "loss": 0.4021, "lr": 9.96960927526972e-06, "epoch": 6.969050160085379, "percentage": 87.11, "elapsed_time": "3:32:10", "remaining_time": "0:31:23"} +{"current_steps": 6535, "total_steps": 7496, "loss": 0.4005, "lr": 9.868503341882872e-06, "epoch": 6.974386339381003, "percentage": 87.18, "elapsed_time": "3:32:48", "remaining_time": "0:31:17"} +{"current_steps": 6540, "total_steps": 7496, "loss": 0.3991, "lr": 9.767886087260125e-06, "epoch": 6.979722518676628, "percentage": 87.25, "elapsed_time": "3:33:25", "remaining_time": "0:31:11"} +{"current_steps": 6545, "total_steps": 7496, "loss": 0.4001, "lr": 9.667758056932341e-06, "epoch": 6.985058697972252, "percentage": 87.31, "elapsed_time": "3:34:05", "remaining_time": "0:31:06"} +{"current_steps": 6550, "total_steps": 7496, "loss": 0.4004, "lr": 9.568119793777952e-06, "epoch": 6.990394877267876, "percentage": 87.38, "elapsed_time": "3:34:40", "remaining_time": "0:31:00"} +{"current_steps": 6555, "total_steps": 7496, "loss": 0.3978, "lr": 9.468971838019869e-06, "epoch": 6.995731056563501, "percentage": 87.45, "elapsed_time": "3:35:14", "remaining_time": "0:30:53"} +{"current_steps": 6560, "total_steps": 7496, "loss": 0.4021, "lr": 9.370314727222718e-06, "epoch": 7.001067235859125, "percentage": 87.51, "elapsed_time": "3:35:51", "remaining_time": "0:30:48"} +{"current_steps": 6565, "total_steps": 7496, "loss": 0.3979, "lr": 9.272148996289765e-06, "epoch": 7.006403415154749, "percentage": 87.58, "elapsed_time": "3:36:29", "remaining_time": "0:30:42"} +{"current_steps": 6570, "total_steps": 7496, "loss": 0.3985, "lr": 9.174475177460096e-06, "epoch": 7.011739594450374, "percentage": 87.65, "elapsed_time": "3:37:05", "remaining_time": "0:30:35"} +{"current_steps": 6575, "total_steps": 7496, "loss": 0.4007, "lr": 9.077293800305842e-06, "epoch": 7.017075773745998, "percentage": 87.71, "elapsed_time": "3:37:40", "remaining_time": "0:30:29"} +{"current_steps": 6580, "total_steps": 7496, "loss": 0.3983, "lr": 8.980605391729035e-06, "epoch": 7.0224119530416225, "percentage": 87.78, "elapsed_time": "3:38:17", "remaining_time": "0:30:23"} +{"current_steps": 6585, "total_steps": 7496, "loss": 0.3979, "lr": 8.884410475959026e-06, "epoch": 7.027748132337247, "percentage": 87.85, "elapsed_time": "3:38:53", "remaining_time": "0:30:16"} +{"current_steps": 6590, "total_steps": 7496, "loss": 0.3964, "lr": 8.788709574549469e-06, "epoch": 7.033084311632871, "percentage": 87.91, "elapsed_time": "3:39:30", "remaining_time": "0:30:10"} +{"current_steps": 6595, "total_steps": 7496, "loss": 0.3989, "lr": 8.693503206375531e-06, "epoch": 7.0384204909284955, "percentage": 87.98, "elapsed_time": "3:40:08", "remaining_time": "0:30:04"} +{"current_steps": 6600, "total_steps": 7496, "loss": 0.3979, "lr": 8.598791887631164e-06, "epoch": 7.04375667022412, "percentage": 88.05, "elapsed_time": "3:40:43", "remaining_time": "0:29:57"} +{"current_steps": 6605, "total_steps": 7496, "loss": 0.3992, "lr": 8.504576131826181e-06, "epoch": 7.049092849519744, "percentage": 88.11, "elapsed_time": "3:41:32", "remaining_time": "0:29:53"} +{"current_steps": 6610, "total_steps": 7496, "loss": 0.3984, "lr": 8.4108564497836e-06, "epoch": 7.054429028815369, "percentage": 88.18, "elapsed_time": "3:42:08", "remaining_time": "0:29:46"} +{"current_steps": 6615, "total_steps": 7496, "loss": 0.3967, "lr": 8.317633349636712e-06, "epoch": 7.059765208110993, "percentage": 88.25, "elapsed_time": "3:42:46", "remaining_time": "0:29:40"} +{"current_steps": 6620, "total_steps": 7496, "loss": 0.3987, "lr": 8.224907336826504e-06, "epoch": 7.065101387406617, "percentage": 88.31, "elapsed_time": "3:43:19", "remaining_time": "0:29:33"} +{"current_steps": 6625, "total_steps": 7496, "loss": 0.3988, "lr": 8.132678914098779e-06, "epoch": 7.070437566702241, "percentage": 88.38, "elapsed_time": "3:43:58", "remaining_time": "0:29:26"} +{"current_steps": 6630, "total_steps": 7496, "loss": 0.3983, "lr": 8.040948581501495e-06, "epoch": 7.075773745997865, "percentage": 88.45, "elapsed_time": "3:44:37", "remaining_time": "0:29:20"} +{"current_steps": 6635, "total_steps": 7496, "loss": 0.3961, "lr": 7.949716836382048e-06, "epoch": 7.0811099252934895, "percentage": 88.51, "elapsed_time": "3:45:12", "remaining_time": "0:29:13"} +{"current_steps": 6640, "total_steps": 7496, "loss": 0.3996, "lr": 7.858984173384598e-06, "epoch": 7.086446104589114, "percentage": 88.58, "elapsed_time": "3:45:50", "remaining_time": "0:29:06"} +{"current_steps": 6645, "total_steps": 7496, "loss": 0.3967, "lr": 7.768751084447357e-06, "epoch": 7.091782283884738, "percentage": 88.65, "elapsed_time": "3:46:25", "remaining_time": "0:28:59"} +{"current_steps": 6650, "total_steps": 7496, "loss": 0.3968, "lr": 7.67901805879987e-06, "epoch": 7.0971184631803625, "percentage": 88.71, "elapsed_time": "3:47:06", "remaining_time": "0:28:53"} +{"current_steps": 6655, "total_steps": 7496, "loss": 0.3981, "lr": 7.589785582960441e-06, "epoch": 7.102454642475987, "percentage": 88.78, "elapsed_time": "3:47:42", "remaining_time": "0:28:46"} +{"current_steps": 6660, "total_steps": 7496, "loss": 0.398, "lr": 7.501054140733521e-06, "epoch": 7.107790821771611, "percentage": 88.85, "elapsed_time": "3:48:18", "remaining_time": "0:28:39"} +{"current_steps": 6665, "total_steps": 7496, "loss": 0.3973, "lr": 7.412824213206981e-06, "epoch": 7.113127001067236, "percentage": 88.91, "elapsed_time": "3:48:55", "remaining_time": "0:28:32"} +{"current_steps": 6670, "total_steps": 7496, "loss": 0.3957, "lr": 7.325096278749599e-06, "epoch": 7.11846318036286, "percentage": 88.98, "elapsed_time": "3:49:33", "remaining_time": "0:28:25"} +{"current_steps": 6675, "total_steps": 7496, "loss": 0.3971, "lr": 7.237870813008362e-06, "epoch": 7.123799359658484, "percentage": 89.05, "elapsed_time": "3:50:10", "remaining_time": "0:28:18"} +{"current_steps": 6680, "total_steps": 7496, "loss": 0.399, "lr": 7.151148288906051e-06, "epoch": 7.129135538954109, "percentage": 89.11, "elapsed_time": "3:50:48", "remaining_time": "0:28:11"} +{"current_steps": 6685, "total_steps": 7496, "loss": 0.3967, "lr": 7.064929176638469e-06, "epoch": 7.134471718249733, "percentage": 89.18, "elapsed_time": "3:51:23", "remaining_time": "0:28:04"} +{"current_steps": 6690, "total_steps": 7496, "loss": 0.3987, "lr": 6.979213943672058e-06, "epoch": 7.139807897545357, "percentage": 89.25, "elapsed_time": "3:52:01", "remaining_time": "0:27:57"} +{"current_steps": 6695, "total_steps": 7496, "loss": 0.3974, "lr": 6.8940030547413044e-06, "epoch": 7.145144076840982, "percentage": 89.31, "elapsed_time": "3:52:40", "remaining_time": "0:27:50"} +{"current_steps": 6700, "total_steps": 7496, "loss": 0.3978, "lr": 6.809296971846213e-06, "epoch": 7.150480256136606, "percentage": 89.38, "elapsed_time": "3:53:17", "remaining_time": "0:27:43"} +{"current_steps": 6705, "total_steps": 7496, "loss": 0.3979, "lr": 6.725096154249822e-06, "epoch": 7.1558164354322304, "percentage": 89.45, "elapsed_time": "3:54:09", "remaining_time": "0:27:37"} +{"current_steps": 6710, "total_steps": 7496, "loss": 0.3968, "lr": 6.6414010584756826e-06, "epoch": 7.161152614727855, "percentage": 89.51, "elapsed_time": "3:54:47", "remaining_time": "0:27:30"} +{"current_steps": 6715, "total_steps": 7496, "loss": 0.3985, "lr": 6.558212138305375e-06, "epoch": 7.166488794023479, "percentage": 89.58, "elapsed_time": "3:55:24", "remaining_time": "0:27:22"} +{"current_steps": 6720, "total_steps": 7496, "loss": 0.3954, "lr": 6.475529844776163e-06, "epoch": 7.1718249733191035, "percentage": 89.65, "elapsed_time": "3:56:03", "remaining_time": "0:27:15"} +{"current_steps": 6725, "total_steps": 7496, "loss": 0.3971, "lr": 6.3933546261783496e-06, "epoch": 7.177161152614728, "percentage": 89.71, "elapsed_time": "3:56:38", "remaining_time": "0:27:07"} +{"current_steps": 6730, "total_steps": 7496, "loss": 0.3986, "lr": 6.3116869280530514e-06, "epoch": 7.182497331910352, "percentage": 89.78, "elapsed_time": "3:57:15", "remaining_time": "0:27:00"} +{"current_steps": 6735, "total_steps": 7496, "loss": 0.3978, "lr": 6.230527193189639e-06, "epoch": 7.187833511205977, "percentage": 89.85, "elapsed_time": "3:57:52", "remaining_time": "0:26:52"} +{"current_steps": 6740, "total_steps": 7496, "loss": 0.3968, "lr": 6.149875861623411e-06, "epoch": 7.193169690501601, "percentage": 89.91, "elapsed_time": "3:58:30", "remaining_time": "0:26:45"} +{"current_steps": 6745, "total_steps": 7496, "loss": 0.3986, "lr": 6.069733370633168e-06, "epoch": 7.198505869797225, "percentage": 89.98, "elapsed_time": "3:59:08", "remaining_time": "0:26:37"} +{"current_steps": 6750, "total_steps": 7496, "loss": 0.3974, "lr": 5.990100154738842e-06, "epoch": 7.20384204909285, "percentage": 90.05, "elapsed_time": "3:59:45", "remaining_time": "0:26:29"} +{"current_steps": 6755, "total_steps": 7496, "loss": 0.3985, "lr": 5.910976645699174e-06, "epoch": 7.209178228388474, "percentage": 90.11, "elapsed_time": "4:00:22", "remaining_time": "0:26:22"} +{"current_steps": 6760, "total_steps": 7496, "loss": 0.3982, "lr": 5.832363272509378e-06, "epoch": 7.214514407684098, "percentage": 90.18, "elapsed_time": "4:00:59", "remaining_time": "0:26:14"} +{"current_steps": 6765, "total_steps": 7496, "loss": 0.3975, "lr": 5.754260461398764e-06, "epoch": 7.219850586979723, "percentage": 90.25, "elapsed_time": "4:01:37", "remaining_time": "0:26:06"} +{"current_steps": 6770, "total_steps": 7496, "loss": 0.399, "lr": 5.6766686358284525e-06, "epoch": 7.225186766275347, "percentage": 90.31, "elapsed_time": "4:02:14", "remaining_time": "0:25:58"} +{"current_steps": 6775, "total_steps": 7496, "loss": 0.4004, "lr": 5.599588216489071e-06, "epoch": 7.230522945570971, "percentage": 90.38, "elapsed_time": "4:02:49", "remaining_time": "0:25:50"} +{"current_steps": 6780, "total_steps": 7496, "loss": 0.3969, "lr": 5.523019621298542e-06, "epoch": 7.235859124866596, "percentage": 90.45, "elapsed_time": "4:03:27", "remaining_time": "0:25:42"} +{"current_steps": 6785, "total_steps": 7496, "loss": 0.3982, "lr": 5.446963265399674e-06, "epoch": 7.24119530416222, "percentage": 90.51, "elapsed_time": "4:04:03", "remaining_time": "0:25:34"} +{"current_steps": 6790, "total_steps": 7496, "loss": 0.3986, "lr": 5.371419561158042e-06, "epoch": 7.2465314834578445, "percentage": 90.58, "elapsed_time": "4:04:38", "remaining_time": "0:25:26"} +{"current_steps": 6795, "total_steps": 7496, "loss": 0.3971, "lr": 5.296388918159723e-06, "epoch": 7.251867662753469, "percentage": 90.65, "elapsed_time": "4:05:15", "remaining_time": "0:25:18"} +{"current_steps": 6800, "total_steps": 7496, "loss": 0.3971, "lr": 5.221871743209039e-06, "epoch": 7.257203842049093, "percentage": 90.72, "elapsed_time": "4:05:54", "remaining_time": "0:25:10"} +{"current_steps": 6805, "total_steps": 7496, "loss": 0.3966, "lr": 5.1478684403263425e-06, "epoch": 7.2625400213447175, "percentage": 90.78, "elapsed_time": "4:06:43", "remaining_time": "0:25:03"} +{"current_steps": 6810, "total_steps": 7496, "loss": 0.3978, "lr": 5.0743794107458664e-06, "epoch": 7.267876200640342, "percentage": 90.85, "elapsed_time": "4:07:20", "remaining_time": "0:24:54"} +{"current_steps": 6815, "total_steps": 7496, "loss": 0.3978, "lr": 5.001405052913577e-06, "epoch": 7.273212379935966, "percentage": 90.92, "elapsed_time": "4:07:57", "remaining_time": "0:24:46"} +{"current_steps": 6820, "total_steps": 7496, "loss": 0.3985, "lr": 4.9289457624848714e-06, "epoch": 7.27854855923159, "percentage": 90.98, "elapsed_time": "4:08:31", "remaining_time": "0:24:38"} +{"current_steps": 6825, "total_steps": 7496, "loss": 0.3953, "lr": 4.85700193232268e-06, "epoch": 7.283884738527215, "percentage": 91.05, "elapsed_time": "4:09:09", "remaining_time": "0:24:29"} +{"current_steps": 6830, "total_steps": 7496, "loss": 0.3986, "lr": 4.785573952495059e-06, "epoch": 7.289220917822838, "percentage": 91.12, "elapsed_time": "4:09:48", "remaining_time": "0:24:21"} +{"current_steps": 6835, "total_steps": 7496, "loss": 0.3998, "lr": 4.714662210273246e-06, "epoch": 7.294557097118463, "percentage": 91.18, "elapsed_time": "4:10:24", "remaining_time": "0:24:12"} +{"current_steps": 6840, "total_steps": 7496, "loss": 0.399, "lr": 4.644267090129562e-06, "epoch": 7.299893276414087, "percentage": 91.25, "elapsed_time": "4:10:59", "remaining_time": "0:24:04"} +{"current_steps": 6845, "total_steps": 7496, "loss": 0.3987, "lr": 4.5743889737352215e-06, "epoch": 7.3052294557097115, "percentage": 91.32, "elapsed_time": "4:11:41", "remaining_time": "0:23:56"} +{"current_steps": 6850, "total_steps": 7496, "loss": 0.3948, "lr": 4.505028239958353e-06, "epoch": 7.310565635005336, "percentage": 91.38, "elapsed_time": "4:12:18", "remaining_time": "0:23:47"} +{"current_steps": 6855, "total_steps": 7496, "loss": 0.3966, "lr": 4.436185264861936e-06, "epoch": 7.31590181430096, "percentage": 91.45, "elapsed_time": "4:12:55", "remaining_time": "0:23:39"} +{"current_steps": 6860, "total_steps": 7496, "loss": 0.3959, "lr": 4.367860421701731e-06, "epoch": 7.321237993596585, "percentage": 91.52, "elapsed_time": "4:13:31", "remaining_time": "0:23:30"} +{"current_steps": 6865, "total_steps": 7496, "loss": 0.3971, "lr": 4.300054080924254e-06, "epoch": 7.326574172892209, "percentage": 91.58, "elapsed_time": "4:14:10", "remaining_time": "0:23:21"} +{"current_steps": 6870, "total_steps": 7496, "loss": 0.3985, "lr": 4.2327666101647735e-06, "epoch": 7.331910352187833, "percentage": 91.65, "elapsed_time": "4:14:46", "remaining_time": "0:23:12"} +{"current_steps": 6875, "total_steps": 7496, "loss": 0.3971, "lr": 4.165998374245395e-06, "epoch": 7.337246531483458, "percentage": 91.72, "elapsed_time": "4:15:22", "remaining_time": "0:23:04"} +{"current_steps": 6880, "total_steps": 7496, "loss": 0.3989, "lr": 4.099749735172909e-06, "epoch": 7.342582710779082, "percentage": 91.78, "elapsed_time": "4:16:00", "remaining_time": "0:22:55"} +{"current_steps": 6885, "total_steps": 7496, "loss": 0.397, "lr": 4.034021052137027e-06, "epoch": 7.347918890074706, "percentage": 91.85, "elapsed_time": "4:16:37", "remaining_time": "0:22:46"} +{"current_steps": 6890, "total_steps": 7496, "loss": 0.3971, "lr": 3.968812681508316e-06, "epoch": 7.353255069370331, "percentage": 91.92, "elapsed_time": "4:17:13", "remaining_time": "0:22:37"} +{"current_steps": 6895, "total_steps": 7496, "loss": 0.3984, "lr": 3.904124976836254e-06, "epoch": 7.358591248665955, "percentage": 91.98, "elapsed_time": "4:17:53", "remaining_time": "0:22:28"} +{"current_steps": 6900, "total_steps": 7496, "loss": 0.3958, "lr": 3.8399582888473895e-06, "epoch": 7.363927427961579, "percentage": 92.05, "elapsed_time": "4:18:31", "remaining_time": "0:22:19"} +{"current_steps": 6905, "total_steps": 7496, "loss": 0.3972, "lr": 3.776312965443374e-06, "epoch": 7.369263607257204, "percentage": 92.12, "elapsed_time": "4:19:15", "remaining_time": "0:22:11"} +{"current_steps": 6910, "total_steps": 7496, "loss": 0.3981, "lr": 3.713189351699131e-06, "epoch": 7.374599786552828, "percentage": 92.18, "elapsed_time": "4:19:54", "remaining_time": "0:22:02"} +{"current_steps": 6915, "total_steps": 7496, "loss": 0.3975, "lr": 3.6505877898609263e-06, "epoch": 7.3799359658484525, "percentage": 92.25, "elapsed_time": "4:20:31", "remaining_time": "0:21:53"} +{"current_steps": 6920, "total_steps": 7496, "loss": 0.3964, "lr": 3.5885086193445883e-06, "epoch": 7.385272145144077, "percentage": 92.32, "elapsed_time": "4:21:06", "remaining_time": "0:21:43"} +{"current_steps": 6925, "total_steps": 7496, "loss": 0.3948, "lr": 3.526952176733578e-06, "epoch": 7.390608324439701, "percentage": 92.38, "elapsed_time": "4:21:43", "remaining_time": "0:21:34"} +{"current_steps": 6930, "total_steps": 7496, "loss": 0.3977, "lr": 3.46591879577719e-06, "epoch": 7.3959445037353255, "percentage": 92.45, "elapsed_time": "4:22:22", "remaining_time": "0:21:25"} +{"current_steps": 6935, "total_steps": 7496, "loss": 0.3965, "lr": 3.4054088073888435e-06, "epoch": 7.40128068303095, "percentage": 92.52, "elapsed_time": "4:22:56", "remaining_time": "0:21:16"} +{"current_steps": 6940, "total_steps": 7496, "loss": 0.3973, "lr": 3.3454225396441164e-06, "epoch": 7.406616862326574, "percentage": 92.58, "elapsed_time": "4:23:33", "remaining_time": "0:21:06"} +{"current_steps": 6945, "total_steps": 7496, "loss": 0.3966, "lr": 3.285960317779102e-06, "epoch": 7.411953041622199, "percentage": 92.65, "elapsed_time": "4:24:13", "remaining_time": "0:20:57"} +{"current_steps": 6950, "total_steps": 7496, "loss": 0.3998, "lr": 3.2270224641886117e-06, "epoch": 7.417289220917823, "percentage": 92.72, "elapsed_time": "4:24:47", "remaining_time": "0:20:48"} +{"current_steps": 6955, "total_steps": 7496, "loss": 0.3941, "lr": 3.168609298424374e-06, "epoch": 7.422625400213447, "percentage": 92.78, "elapsed_time": "4:25:23", "remaining_time": "0:20:38"} +{"current_steps": 6960, "total_steps": 7496, "loss": 0.397, "lr": 3.110721137193384e-06, "epoch": 7.427961579509072, "percentage": 92.85, "elapsed_time": "4:26:01", "remaining_time": "0:20:29"} +{"current_steps": 6965, "total_steps": 7496, "loss": 0.3953, "lr": 3.053358294356101e-06, "epoch": 7.433297758804696, "percentage": 92.92, "elapsed_time": "4:26:37", "remaining_time": "0:20:19"} +{"current_steps": 6970, "total_steps": 7496, "loss": 0.3988, "lr": 2.9965210809248412e-06, "epoch": 7.43863393810032, "percentage": 92.98, "elapsed_time": "4:27:13", "remaining_time": "0:20:09"} +{"current_steps": 6975, "total_steps": 7496, "loss": 0.3965, "lr": 2.9402098050619775e-06, "epoch": 7.443970117395945, "percentage": 93.05, "elapsed_time": "4:27:51", "remaining_time": "0:20:00"} +{"current_steps": 6980, "total_steps": 7496, "loss": 0.3966, "lr": 2.884424772078398e-06, "epoch": 7.449306296691569, "percentage": 93.12, "elapsed_time": "4:28:29", "remaining_time": "0:19:50"} +{"current_steps": 6985, "total_steps": 7496, "loss": 0.3992, "lr": 2.829166284431761e-06, "epoch": 7.454642475987193, "percentage": 93.18, "elapsed_time": "4:29:05", "remaining_time": "0:19:41"} +{"current_steps": 6990, "total_steps": 7496, "loss": 0.3974, "lr": 2.7744346417248546e-06, "epoch": 7.459978655282818, "percentage": 93.25, "elapsed_time": "4:29:41", "remaining_time": "0:19:31"} +{"current_steps": 6995, "total_steps": 7496, "loss": 0.3966, "lr": 2.7202301407040274e-06, "epoch": 7.465314834578442, "percentage": 93.32, "elapsed_time": "4:30:20", "remaining_time": "0:19:21"} +{"current_steps": 7000, "total_steps": 7496, "loss": 0.3965, "lr": 2.6665530752575274e-06, "epoch": 7.4706510138740665, "percentage": 93.38, "elapsed_time": "4:30:56", "remaining_time": "0:19:11"} +{"current_steps": 7005, "total_steps": 7496, "loss": 0.3982, "lr": 2.613403736413944e-06, "epoch": 7.475987193169691, "percentage": 93.45, "elapsed_time": "4:31:45", "remaining_time": "0:19:02"} +{"current_steps": 7010, "total_steps": 7496, "loss": 0.398, "lr": 2.560782412340623e-06, "epoch": 7.481323372465315, "percentage": 93.52, "elapsed_time": "4:32:26", "remaining_time": "0:18:53"} +{"current_steps": 7015, "total_steps": 7496, "loss": 0.3973, "lr": 2.508689388342078e-06, "epoch": 7.48665955176094, "percentage": 93.58, "elapsed_time": "4:33:00", "remaining_time": "0:18:43"} +{"current_steps": 7020, "total_steps": 7496, "loss": 0.3977, "lr": 2.45712494685848e-06, "epoch": 7.491995731056564, "percentage": 93.65, "elapsed_time": "4:33:35", "remaining_time": "0:18:33"} +{"current_steps": 7025, "total_steps": 7496, "loss": 0.398, "lr": 2.4060893674640815e-06, "epoch": 7.497331910352187, "percentage": 93.72, "elapsed_time": "4:34:13", "remaining_time": "0:18:23"} +{"current_steps": 7030, "total_steps": 7496, "loss": 0.398, "lr": 2.3555829268657626e-06, "epoch": 7.502668089647813, "percentage": 93.78, "elapsed_time": "4:34:53", "remaining_time": "0:18:13"} +{"current_steps": 7035, "total_steps": 7496, "loss": 0.3979, "lr": 2.305605898901486e-06, "epoch": 7.508004268943436, "percentage": 93.85, "elapsed_time": "4:35:26", "remaining_time": "0:18:02"} +{"current_steps": 7040, "total_steps": 7496, "loss": 0.3971, "lr": 2.25615855453879e-06, "epoch": 7.513340448239061, "percentage": 93.92, "elapsed_time": "4:36:04", "remaining_time": "0:17:52"} +{"current_steps": 7045, "total_steps": 7496, "loss": 0.3966, "lr": 2.207241161873419e-06, "epoch": 7.518676627534685, "percentage": 93.98, "elapsed_time": "4:36:42", "remaining_time": "0:17:42"} +{"current_steps": 7050, "total_steps": 7496, "loss": 0.3943, "lr": 2.15885398612774e-06, "epoch": 7.524012806830309, "percentage": 94.05, "elapsed_time": "4:37:19", "remaining_time": "0:17:32"} +{"current_steps": 7055, "total_steps": 7496, "loss": 0.3965, "lr": 2.110997289649397e-06, "epoch": 7.5293489861259335, "percentage": 94.12, "elapsed_time": "4:37:56", "remaining_time": "0:17:22"} +{"current_steps": 7060, "total_steps": 7496, "loss": 0.3978, "lr": 2.0636713319098444e-06, "epoch": 7.534685165421558, "percentage": 94.18, "elapsed_time": "4:38:35", "remaining_time": "0:17:12"} +{"current_steps": 7065, "total_steps": 7496, "loss": 0.3971, "lr": 2.0168763695029735e-06, "epoch": 7.540021344717182, "percentage": 94.25, "elapsed_time": "4:39:09", "remaining_time": "0:17:01"} +{"current_steps": 7070, "total_steps": 7496, "loss": 0.3993, "lr": 1.9706126561436775e-06, "epoch": 7.545357524012807, "percentage": 94.32, "elapsed_time": "4:39:46", "remaining_time": "0:16:51"} +{"current_steps": 7075, "total_steps": 7496, "loss": 0.3963, "lr": 1.9248804426665323e-06, "epoch": 7.550693703308431, "percentage": 94.38, "elapsed_time": "4:40:26", "remaining_time": "0:16:41"} +{"current_steps": 7080, "total_steps": 7496, "loss": 0.3963, "lr": 1.8796799770243845e-06, "epoch": 7.556029882604055, "percentage": 94.45, "elapsed_time": "4:41:00", "remaining_time": "0:16:30"} +{"current_steps": 7085, "total_steps": 7496, "loss": 0.3965, "lr": 1.835011504287032e-06, "epoch": 7.56136606189968, "percentage": 94.52, "elapsed_time": "4:41:36", "remaining_time": "0:16:20"} +{"current_steps": 7090, "total_steps": 7496, "loss": 0.3965, "lr": 1.790875266639891e-06, "epoch": 7.566702241195304, "percentage": 94.58, "elapsed_time": "4:42:15", "remaining_time": "0:16:09"} +{"current_steps": 7095, "total_steps": 7496, "loss": 0.396, "lr": 1.7472715033826747e-06, "epoch": 7.572038420490928, "percentage": 94.65, "elapsed_time": "4:42:50", "remaining_time": "0:15:59"} +{"current_steps": 7100, "total_steps": 7496, "loss": 0.3985, "lr": 1.7042004509281284e-06, "epoch": 7.577374599786553, "percentage": 94.72, "elapsed_time": "4:43:24", "remaining_time": "0:15:48"} +{"current_steps": 7105, "total_steps": 7496, "loss": 0.3971, "lr": 1.6616623428006961e-06, "epoch": 7.582710779082177, "percentage": 94.78, "elapsed_time": "4:44:15", "remaining_time": "0:15:38"} +{"current_steps": 7110, "total_steps": 7496, "loss": 0.397, "lr": 1.6196574096353e-06, "epoch": 7.588046958377801, "percentage": 94.85, "elapsed_time": "4:44:50", "remaining_time": "0:15:27"} +{"current_steps": 7115, "total_steps": 7496, "loss": 0.3971, "lr": 1.578185879176064e-06, "epoch": 7.593383137673426, "percentage": 94.92, "elapsed_time": "4:45:24", "remaining_time": "0:15:16"} +{"current_steps": 7120, "total_steps": 7496, "loss": 0.3978, "lr": 1.5372479762750691e-06, "epoch": 7.59871931696905, "percentage": 94.98, "elapsed_time": "4:46:02", "remaining_time": "0:15:06"} +{"current_steps": 7125, "total_steps": 7496, "loss": 0.3983, "lr": 1.4968439228912e-06, "epoch": 7.6040554962646745, "percentage": 95.05, "elapsed_time": "4:46:38", "remaining_time": "0:14:55"} +{"current_steps": 7130, "total_steps": 7496, "loss": 0.3972, "lr": 1.4569739380888458e-06, "epoch": 7.609391675560299, "percentage": 95.12, "elapsed_time": "4:47:12", "remaining_time": "0:14:44"} +{"current_steps": 7135, "total_steps": 7496, "loss": 0.3994, "lr": 1.4176382380367782e-06, "epoch": 7.614727854855923, "percentage": 95.18, "elapsed_time": "4:47:48", "remaining_time": "0:14:33"} +{"current_steps": 7140, "total_steps": 7496, "loss": 0.399, "lr": 1.3788370360069636e-06, "epoch": 7.6200640341515475, "percentage": 95.25, "elapsed_time": "4:48:28", "remaining_time": "0:14:22"} +{"current_steps": 7145, "total_steps": 7496, "loss": 0.396, "lr": 1.3405705423734094e-06, "epoch": 7.625400213447172, "percentage": 95.32, "elapsed_time": "4:49:02", "remaining_time": "0:14:11"} +{"current_steps": 7150, "total_steps": 7496, "loss": 0.397, "lr": 1.3028389646110084e-06, "epoch": 7.630736392742796, "percentage": 95.38, "elapsed_time": "4:49:39", "remaining_time": "0:14:01"} +{"current_steps": 7155, "total_steps": 7496, "loss": 0.3979, "lr": 1.2656425072944068e-06, "epoch": 7.636072572038421, "percentage": 95.45, "elapsed_time": "4:50:16", "remaining_time": "0:13:50"} +{"current_steps": 7160, "total_steps": 7496, "loss": 0.3954, "lr": 1.228981372096949e-06, "epoch": 7.641408751334045, "percentage": 95.52, "elapsed_time": "4:50:49", "remaining_time": "0:13:38"} +{"current_steps": 7165, "total_steps": 7496, "loss": 0.396, "lr": 1.192855757789546e-06, "epoch": 7.646744930629669, "percentage": 95.58, "elapsed_time": "4:51:27", "remaining_time": "0:13:27"} +{"current_steps": 7170, "total_steps": 7496, "loss": 0.3987, "lr": 1.1572658602395426e-06, "epoch": 7.652081109925294, "percentage": 95.65, "elapsed_time": "4:52:04", "remaining_time": "0:13:16"} +{"current_steps": 7175, "total_steps": 7496, "loss": 0.3987, "lr": 1.122211872409784e-06, "epoch": 7.657417289220918, "percentage": 95.72, "elapsed_time": "4:52:40", "remaining_time": "0:13:05"} +{"current_steps": 7180, "total_steps": 7496, "loss": 0.3985, "lr": 1.087693984357452e-06, "epoch": 7.662753468516542, "percentage": 95.78, "elapsed_time": "4:53:16", "remaining_time": "0:12:54"} +{"current_steps": 7185, "total_steps": 7496, "loss": 0.3976, "lr": 1.0537123832330964e-06, "epoch": 7.668089647812167, "percentage": 95.85, "elapsed_time": "4:53:53", "remaining_time": "0:12:43"} +{"current_steps": 7190, "total_steps": 7496, "loss": 0.3977, "lr": 1.0202672532796055e-06, "epoch": 7.673425827107791, "percentage": 95.92, "elapsed_time": "4:54:29", "remaining_time": "0:12:31"} +{"current_steps": 7195, "total_steps": 7496, "loss": 0.3971, "lr": 9.87358775831182e-07, "epoch": 7.678762006403415, "percentage": 95.98, "elapsed_time": "4:55:03", "remaining_time": "0:12:20"} +{"current_steps": 7200, "total_steps": 7496, "loss": 0.397, "lr": 9.549871293124234e-07, "epoch": 7.68409818569904, "percentage": 96.05, "elapsed_time": "4:55:40", "remaining_time": "0:12:09"} +{"current_steps": 7205, "total_steps": 7496, "loss": 0.3971, "lr": 9.231524892372889e-07, "epoch": 7.689434364994664, "percentage": 96.12, "elapsed_time": "4:56:30", "remaining_time": "0:11:58"} +{"current_steps": 7210, "total_steps": 7496, "loss": 0.399, "lr": 8.918550282081772e-07, "epoch": 7.6947705442902885, "percentage": 96.18, "elapsed_time": "4:57:04", "remaining_time": "0:11:47"} +{"current_steps": 7215, "total_steps": 7496, "loss": 0.3961, "lr": 8.610949159149728e-07, "epoch": 7.700106723585913, "percentage": 96.25, "elapsed_time": "4:57:42", "remaining_time": "0:11:35"} +{"current_steps": 7220, "total_steps": 7496, "loss": 0.3967, "lr": 8.308723191341683e-07, "epoch": 7.705442902881536, "percentage": 96.32, "elapsed_time": "4:58:21", "remaining_time": "0:11:24"} +{"current_steps": 7225, "total_steps": 7496, "loss": 0.3972, "lr": 8.011874017279208e-07, "epoch": 7.710779082177162, "percentage": 96.38, "elapsed_time": "4:58:56", "remaining_time": "0:11:12"} +{"current_steps": 7230, "total_steps": 7496, "loss": 0.3977, "lr": 7.72040324643164e-07, "epoch": 7.716115261472785, "percentage": 96.45, "elapsed_time": "4:59:32", "remaining_time": "0:11:01"} +{"current_steps": 7235, "total_steps": 7496, "loss": 0.3994, "lr": 7.434312459107639e-07, "epoch": 7.72145144076841, "percentage": 96.52, "elapsed_time": "5:00:08", "remaining_time": "0:10:49"} +{"current_steps": 7240, "total_steps": 7496, "loss": 0.3975, "lr": 7.153603206446203e-07, "epoch": 7.726787620064034, "percentage": 96.58, "elapsed_time": "5:00:46", "remaining_time": "0:10:38"} +{"current_steps": 7245, "total_steps": 7496, "loss": 0.3963, "lr": 6.878277010408774e-07, "epoch": 7.732123799359658, "percentage": 96.65, "elapsed_time": "5:01:23", "remaining_time": "0:10:26"} +{"current_steps": 7250, "total_steps": 7496, "loss": 0.3975, "lr": 6.60833536377059e-07, "epoch": 7.7374599786552825, "percentage": 96.72, "elapsed_time": "5:02:00", "remaining_time": "0:10:14"} +{"current_steps": 7255, "total_steps": 7496, "loss": 0.3973, "lr": 6.343779730112464e-07, "epoch": 7.742796157950907, "percentage": 96.78, "elapsed_time": "5:02:37", "remaining_time": "0:10:03"} +{"current_steps": 7260, "total_steps": 7496, "loss": 0.3966, "lr": 6.084611543813679e-07, "epoch": 7.748132337246531, "percentage": 96.85, "elapsed_time": "5:03:14", "remaining_time": "0:09:51"} +{"current_steps": 7265, "total_steps": 7496, "loss": 0.396, "lr": 5.830832210042991e-07, "epoch": 7.7534685165421555, "percentage": 96.92, "elapsed_time": "5:03:52", "remaining_time": "0:09:39"} +{"current_steps": 7270, "total_steps": 7496, "loss": 0.3966, "lr": 5.582443104752089e-07, "epoch": 7.75880469583778, "percentage": 96.99, "elapsed_time": "5:04:30", "remaining_time": "0:09:27"} +{"current_steps": 7275, "total_steps": 7496, "loss": 0.3981, "lr": 5.339445574667479e-07, "epoch": 7.764140875133404, "percentage": 97.05, "elapsed_time": "5:05:05", "remaining_time": "0:09:16"} +{"current_steps": 7280, "total_steps": 7496, "loss": 0.3985, "lr": 5.101840937283498e-07, "epoch": 7.769477054429029, "percentage": 97.12, "elapsed_time": "5:05:42", "remaining_time": "0:09:04"} +{"current_steps": 7285, "total_steps": 7496, "loss": 0.3966, "lr": 4.869630480854981e-07, "epoch": 7.774813233724653, "percentage": 97.19, "elapsed_time": "5:06:19", "remaining_time": "0:08:52"} +{"current_steps": 7290, "total_steps": 7496, "loss": 0.399, "lr": 4.6428154643904933e-07, "epoch": 7.780149413020277, "percentage": 97.25, "elapsed_time": "5:06:55", "remaining_time": "0:08:40"} +{"current_steps": 7295, "total_steps": 7496, "loss": 0.3938, "lr": 4.4213971176451095e-07, "epoch": 7.785485592315902, "percentage": 97.32, "elapsed_time": "5:07:31", "remaining_time": "0:08:28"} +{"current_steps": 7300, "total_steps": 7496, "loss": 0.3985, "lr": 4.2053766411144225e-07, "epoch": 7.790821771611526, "percentage": 97.39, "elapsed_time": "5:08:09", "remaining_time": "0:08:16"} +{"current_steps": 7305, "total_steps": 7496, "loss": 0.3956, "lr": 3.994755206027212e-07, "epoch": 7.79615795090715, "percentage": 97.45, "elapsed_time": "5:08:58", "remaining_time": "0:08:04"} +{"current_steps": 7310, "total_steps": 7496, "loss": 0.3983, "lr": 3.789533954339897e-07, "epoch": 7.801494130202775, "percentage": 97.52, "elapsed_time": "5:09:36", "remaining_time": "0:07:52"} +{"current_steps": 7315, "total_steps": 7496, "loss": 0.397, "lr": 3.589713998729427e-07, "epoch": 7.806830309498399, "percentage": 97.59, "elapsed_time": "5:10:11", "remaining_time": "0:07:40"} +{"current_steps": 7320, "total_steps": 7496, "loss": 0.3965, "lr": 3.395296422588512e-07, "epoch": 7.812166488794023, "percentage": 97.65, "elapsed_time": "5:10:48", "remaining_time": "0:07:28"} +{"current_steps": 7325, "total_steps": 7496, "loss": 0.3974, "lr": 3.20628228001818e-07, "epoch": 7.817502668089648, "percentage": 97.72, "elapsed_time": "5:11:27", "remaining_time": "0:07:16"} +{"current_steps": 7330, "total_steps": 7496, "loss": 0.3966, "lr": 3.022672595823672e-07, "epoch": 7.822838847385272, "percentage": 97.79, "elapsed_time": "5:12:02", "remaining_time": "0:07:03"} +{"current_steps": 7335, "total_steps": 7496, "loss": 0.3978, "lr": 2.844468365507447e-07, "epoch": 7.8281750266808965, "percentage": 97.85, "elapsed_time": "5:12:37", "remaining_time": "0:06:51"} +{"current_steps": 7340, "total_steps": 7496, "loss": 0.3979, "lr": 2.6716705552649603e-07, "epoch": 7.833511205976521, "percentage": 97.92, "elapsed_time": "5:13:15", "remaining_time": "0:06:39"} +{"current_steps": 7345, "total_steps": 7496, "loss": 0.398, "lr": 2.5042801019785624e-07, "epoch": 7.838847385272145, "percentage": 97.99, "elapsed_time": "5:13:51", "remaining_time": "0:06:27"} +{"current_steps": 7350, "total_steps": 7496, "loss": 0.3957, "lr": 2.342297913212943e-07, "epoch": 7.8441835645677696, "percentage": 98.05, "elapsed_time": "5:14:27", "remaining_time": "0:06:14"} +{"current_steps": 7355, "total_steps": 7496, "loss": 0.395, "lr": 2.1857248672100262e-07, "epoch": 7.849519743863394, "percentage": 98.12, "elapsed_time": "5:15:05", "remaining_time": "0:06:02"} +{"current_steps": 7360, "total_steps": 7496, "loss": 0.3971, "lr": 2.0345618128840838e-07, "epoch": 7.854855923159018, "percentage": 98.19, "elapsed_time": "5:15:41", "remaining_time": "0:05:50"} +{"current_steps": 7365, "total_steps": 7496, "loss": 0.3976, "lr": 1.888809569817518e-07, "epoch": 7.860192102454643, "percentage": 98.25, "elapsed_time": "5:16:16", "remaining_time": "0:05:37"} +{"current_steps": 7370, "total_steps": 7496, "loss": 0.3971, "lr": 1.7484689282558643e-07, "epoch": 7.865528281750267, "percentage": 98.32, "elapsed_time": "5:16:53", "remaining_time": "0:05:25"} +{"current_steps": 7375, "total_steps": 7496, "loss": 0.3973, "lr": 1.6135406491041282e-07, "epoch": 7.870864461045891, "percentage": 98.39, "elapsed_time": "5:17:28", "remaining_time": "0:05:12"} +{"current_steps": 7380, "total_steps": 7496, "loss": 0.397, "lr": 1.484025463921901e-07, "epoch": 7.876200640341516, "percentage": 98.45, "elapsed_time": "5:18:06", "remaining_time": "0:05:00"} +{"current_steps": 7385, "total_steps": 7496, "loss": 0.3991, "lr": 1.359924074920249e-07, "epoch": 7.88153681963714, "percentage": 98.52, "elapsed_time": "5:18:44", "remaining_time": "0:04:47"} +{"current_steps": 7390, "total_steps": 7496, "loss": 0.3992, "lr": 1.2412371549573864e-07, "epoch": 7.886872998932764, "percentage": 98.59, "elapsed_time": "5:19:18", "remaining_time": "0:04:34"} +{"current_steps": 7395, "total_steps": 7496, "loss": 0.3979, "lr": 1.1279653475350094e-07, "epoch": 7.892209178228389, "percentage": 98.65, "elapsed_time": "5:19:58", "remaining_time": "0:04:22"} +{"current_steps": 7400, "total_steps": 7496, "loss": 0.3975, "lr": 1.0201092667950773e-07, "epoch": 7.897545357524013, "percentage": 98.72, "elapsed_time": "5:20:34", "remaining_time": "0:04:09"} +{"current_steps": 7405, "total_steps": 7496, "loss": 0.3983, "lr": 9.176694975161493e-08, "epoch": 7.9028815368196375, "percentage": 98.79, "elapsed_time": "5:21:22", "remaining_time": "0:03:56"} +{"current_steps": 7410, "total_steps": 7496, "loss": 0.3958, "lr": 8.206465951104969e-08, "epoch": 7.908217716115262, "percentage": 98.85, "elapsed_time": "5:21:58", "remaining_time": "0:03:44"} +{"current_steps": 7415, "total_steps": 7496, "loss": 0.3958, "lr": 7.290410856211071e-08, "epoch": 7.913553895410886, "percentage": 98.92, "elapsed_time": "5:22:34", "remaining_time": "0:03:31"} +{"current_steps": 7420, "total_steps": 7496, "loss": 0.397, "lr": 6.428534657185736e-08, "epoch": 7.9188900747065105, "percentage": 98.99, "elapsed_time": "5:23:12", "remaining_time": "0:03:18"} +{"current_steps": 7425, "total_steps": 7496, "loss": 0.3954, "lr": 5.620842026985429e-08, "epoch": 7.924226254002134, "percentage": 99.05, "elapsed_time": "5:23:46", "remaining_time": "0:03:05"} +{"current_steps": 7430, "total_steps": 7496, "loss": 0.3979, "lr": 4.8673373447916116e-08, "epoch": 7.929562433297759, "percentage": 99.12, "elapsed_time": "5:24:26", "remaining_time": "0:02:52"} +{"current_steps": 7435, "total_steps": 7496, "loss": 0.3971, "lr": 4.1680246959896473e-08, "epoch": 7.934898612593383, "percentage": 99.19, "elapsed_time": "5:25:03", "remaining_time": "0:02:40"} +{"current_steps": 7440, "total_steps": 7496, "loss": 0.3968, "lr": 3.522907872141046e-08, "epoch": 7.940234791889008, "percentage": 99.25, "elapsed_time": "5:25:38", "remaining_time": "0:02:27"} +{"current_steps": 7445, "total_steps": 7496, "loss": 0.3956, "lr": 2.9319903709679186e-08, "epoch": 7.945570971184631, "percentage": 99.32, "elapsed_time": "5:26:16", "remaining_time": "0:02:14"} +{"current_steps": 7450, "total_steps": 7496, "loss": 0.3956, "lr": 2.3952753963318865e-08, "epoch": 7.950907150480256, "percentage": 99.39, "elapsed_time": "5:26:54", "remaining_time": "0:02:01"} +{"current_steps": 7455, "total_steps": 7496, "loss": 0.3965, "lr": 1.9127658582163145e-08, "epoch": 7.95624332977588, "percentage": 99.45, "elapsed_time": "5:27:28", "remaining_time": "0:01:48"} +{"current_steps": 7460, "total_steps": 7496, "loss": 0.3969, "lr": 1.4844643727129903e-08, "epoch": 7.9615795090715045, "percentage": 99.52, "elapsed_time": "5:28:03", "remaining_time": "0:01:34"} +{"current_steps": 7465, "total_steps": 7496, "loss": 0.3978, "lr": 1.1103732620043606e-08, "epoch": 7.966915688367129, "percentage": 99.59, "elapsed_time": "5:28:44", "remaining_time": "0:01:21"} +{"current_steps": 7470, "total_steps": 7496, "loss": 0.3963, "lr": 7.90494554353538e-09, "epoch": 7.972251867662753, "percentage": 99.65, "elapsed_time": "5:29:19", "remaining_time": "0:01:08"} +{"current_steps": 7475, "total_steps": 7496, "loss": 0.3993, "lr": 5.248299840920901e-09, "epoch": 7.9775880469583775, "percentage": 99.72, "elapsed_time": "5:29:55", "remaining_time": "0:00:55"} +{"current_steps": 7480, "total_steps": 7496, "loss": 0.3982, "lr": 3.1338099161226653e-09, "epoch": 7.982924226254002, "percentage": 99.79, "elapsed_time": "5:30:33", "remaining_time": "0:00:42"} +{"current_steps": 7485, "total_steps": 7496, "loss": 0.3965, "lr": 1.5614872335589781e-09, "epoch": 7.988260405549626, "percentage": 99.85, "elapsed_time": "5:31:08", "remaining_time": "0:00:29"} +{"current_steps": 7490, "total_steps": 7496, "loss": 0.3989, "lr": 5.313403181328447e-10, "epoch": 7.993596584845251, "percentage": 99.92, "elapsed_time": "5:31:44", "remaining_time": "0:00:15"} +{"current_steps": 7495, "total_steps": 7496, "loss": 0.3961, "lr": 4.3374755120950685e-11, "epoch": 7.998932764140875, "percentage": 99.99, "elapsed_time": "5:32:20", "remaining_time": "0:00:02"} +{"current_steps": 7496, "total_steps": 7496, "epoch": 8.0, "percentage": 100.0, "elapsed_time": "5:32:39", "remaining_time": "0:00:00"} diff --git a/trainer_state.json b/trainer_state.json new file mode 100644 index 0000000..595c099 --- /dev/null +++ b/trainer_state.json @@ -0,0 +1,16536 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.0, + "eval_steps": 500, + "global_step": 7496, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.005336179295624333, + "grad_norm": 12.46944717625873, + "learning_rate": 1.0666666666666667e-06, + "loss": 0.8061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.8061428070068359, + "step": 5, + "valid_targets_mean": 28904.8, + "valid_targets_min": 20949 + }, + { + "epoch": 0.010672358591248666, + "grad_norm": 10.854200136832457, + "learning_rate": 2.4000000000000003e-06, + "loss": 0.7944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7890393137931824, + "step": 10, + "valid_targets_mean": 28786.1, + "valid_targets_min": 20945 + }, + { + "epoch": 0.016008537886873, + "grad_norm": 6.848611631904356, + "learning_rate": 3.7333333333333337e-06, + "loss": 0.7523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7329467535018921, + "step": 15, + "valid_targets_mean": 28843.0, + "valid_targets_min": 20873 + }, + { + "epoch": 0.021344717182497332, + "grad_norm": 2.5190437684250737, + "learning_rate": 5.066666666666667e-06, + "loss": 0.7026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6821247935295105, + "step": 20, + "valid_targets_mean": 28758.3, + "valid_targets_min": 19258 + }, + { + "epoch": 0.026680896478121666, + "grad_norm": 1.447362426625492, + "learning_rate": 6.4000000000000006e-06, + "loss": 0.6692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.665269672870636, + "step": 25, + "valid_targets_mean": 28746.4, + "valid_targets_min": 20839 + }, + { + "epoch": 0.032017075773746, + "grad_norm": 1.3783494582439455, + "learning_rate": 7.733333333333334e-06, + "loss": 0.6579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6507786512374878, + "step": 30, + "valid_targets_mean": 28955.7, + "valid_targets_min": 22126 + }, + { + "epoch": 0.03735325506937033, + "grad_norm": 0.5967416140332733, + "learning_rate": 9.066666666666667e-06, + "loss": 0.6443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6362227201461792, + "step": 35, + "valid_targets_mean": 28837.7, + "valid_targets_min": 22221 + }, + { + "epoch": 0.042689434364994665, + "grad_norm": 0.554654624115749, + "learning_rate": 1.04e-05, + "loss": 0.6329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.631878137588501, + "step": 40, + "valid_targets_mean": 28880.9, + "valid_targets_min": 21716 + }, + { + "epoch": 0.048025613660619, + "grad_norm": 0.40179224460616403, + "learning_rate": 1.1733333333333333e-05, + "loss": 0.6266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6223902702331543, + "step": 45, + "valid_targets_mean": 28925.6, + "valid_targets_min": 19265 + }, + { + "epoch": 0.05336179295624333, + "grad_norm": 0.3137447532636136, + "learning_rate": 1.3066666666666666e-05, + "loss": 0.6208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6185448169708252, + "step": 50, + "valid_targets_mean": 28986.4, + "valid_targets_min": 21522 + }, + { + "epoch": 0.05869797225186766, + "grad_norm": 0.24068546725659087, + "learning_rate": 1.44e-05, + "loss": 0.6136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6145120859146118, + "step": 55, + "valid_targets_mean": 28725.1, + "valid_targets_min": 19987 + }, + { + "epoch": 0.064034151547492, + "grad_norm": 0.22089166210741729, + "learning_rate": 1.5733333333333334e-05, + "loss": 0.6083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6090884208679199, + "step": 60, + "valid_targets_mean": 28757.6, + "valid_targets_min": 20837 + }, + { + "epoch": 0.06937033084311633, + "grad_norm": 0.21982402489152278, + "learning_rate": 1.7066666666666667e-05, + "loss": 0.607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6048812866210938, + "step": 65, + "valid_targets_mean": 28755.8, + "valid_targets_min": 17393 + }, + { + "epoch": 0.07470651013874066, + "grad_norm": 0.2081277489820844, + "learning_rate": 1.84e-05, + "loss": 0.6034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5989037752151489, + "step": 70, + "valid_targets_mean": 28790.1, + "valid_targets_min": 19897 + }, + { + "epoch": 0.08004268943436499, + "grad_norm": 0.22176638943335583, + "learning_rate": 1.9733333333333333e-05, + "loss": 0.5973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5971338748931885, + "step": 75, + "valid_targets_mean": 28797.2, + "valid_targets_min": 20862 + }, + { + "epoch": 0.08537886872998933, + "grad_norm": 0.21267940944215732, + "learning_rate": 2.106666666666667e-05, + "loss": 0.5912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.587620735168457, + "step": 80, + "valid_targets_mean": 28717.8, + "valid_targets_min": 18978 + }, + { + "epoch": 0.09071504802561366, + "grad_norm": 0.2359193466407616, + "learning_rate": 2.2400000000000002e-05, + "loss": 0.5886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.583030641078949, + "step": 85, + "valid_targets_mean": 28823.8, + "valid_targets_min": 17944 + }, + { + "epoch": 0.096051227321238, + "grad_norm": 0.2670343955859104, + "learning_rate": 2.3733333333333335e-05, + "loss": 0.5903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5849412083625793, + "step": 90, + "valid_targets_mean": 28909.0, + "valid_targets_min": 21779 + }, + { + "epoch": 0.10138740661686232, + "grad_norm": 0.2670701546160268, + "learning_rate": 2.5066666666666665e-05, + "loss": 0.5842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5862775444984436, + "step": 95, + "valid_targets_mean": 28871.7, + "valid_targets_min": 22223 + }, + { + "epoch": 0.10672358591248667, + "grad_norm": 0.36156840702341764, + "learning_rate": 2.64e-05, + "loss": 0.5839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5871520042419434, + "step": 100, + "valid_targets_mean": 28859.4, + "valid_targets_min": 19968 + }, + { + "epoch": 0.11205976520811099, + "grad_norm": 0.30448811630120975, + "learning_rate": 2.7733333333333334e-05, + "loss": 0.5835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5811035633087158, + "step": 105, + "valid_targets_mean": 28989.8, + "valid_targets_min": 18150 + }, + { + "epoch": 0.11739594450373532, + "grad_norm": 0.37432798390739574, + "learning_rate": 2.906666666666667e-05, + "loss": 0.5798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5788925886154175, + "step": 110, + "valid_targets_mean": 28699.6, + "valid_targets_min": 21089 + }, + { + "epoch": 0.12273212379935966, + "grad_norm": 0.32787186906252047, + "learning_rate": 3.04e-05, + "loss": 0.5784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5760443210601807, + "step": 115, + "valid_targets_mean": 28889.4, + "valid_targets_min": 21459 + }, + { + "epoch": 0.128068303094984, + "grad_norm": 0.270138890246249, + "learning_rate": 3.173333333333334e-05, + "loss": 0.5753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5727493166923523, + "step": 120, + "valid_targets_mean": 28960.6, + "valid_targets_min": 22036 + }, + { + "epoch": 0.13340448239060831, + "grad_norm": 0.5654127135981639, + "learning_rate": 3.3066666666666666e-05, + "loss": 0.5753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5761654376983643, + "step": 125, + "valid_targets_mean": 28920.7, + "valid_targets_min": 22329 + }, + { + "epoch": 0.13874066168623267, + "grad_norm": 0.5017381861949305, + "learning_rate": 3.4399999999999996e-05, + "loss": 0.5761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5769978761672974, + "step": 130, + "valid_targets_mean": 28861.9, + "valid_targets_min": 16348 + }, + { + "epoch": 0.144076840981857, + "grad_norm": 0.6326271596195411, + "learning_rate": 3.573333333333333e-05, + "loss": 0.576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5722428560256958, + "step": 135, + "valid_targets_mean": 28751.1, + "valid_targets_min": 21044 + }, + { + "epoch": 0.14941302027748132, + "grad_norm": 0.4278862899999183, + "learning_rate": 3.706666666666667e-05, + "loss": 0.5715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5662323236465454, + "step": 140, + "valid_targets_mean": 28761.6, + "valid_targets_min": 21030 + }, + { + "epoch": 0.15474919957310565, + "grad_norm": 0.6394771886602794, + "learning_rate": 3.8400000000000005e-05, + "loss": 0.5701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5686701536178589, + "step": 145, + "valid_targets_mean": 28730.7, + "valid_targets_min": 20582 + }, + { + "epoch": 0.16008537886872998, + "grad_norm": 0.48399755814562095, + "learning_rate": 3.9733333333333335e-05, + "loss": 0.5685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5691660642623901, + "step": 150, + "valid_targets_mean": 28573.6, + "valid_targets_min": 19330 + }, + { + "epoch": 0.16542155816435433, + "grad_norm": 0.661411567386734, + "learning_rate": 4.106666666666667e-05, + "loss": 0.5692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5699338912963867, + "step": 155, + "valid_targets_mean": 28886.6, + "valid_targets_min": 21972 + }, + { + "epoch": 0.17075773745997866, + "grad_norm": 0.5829717337815095, + "learning_rate": 4.24e-05, + "loss": 0.5663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5658145546913147, + "step": 160, + "valid_targets_mean": 28911.6, + "valid_targets_min": 22554 + }, + { + "epoch": 0.17609391675560299, + "grad_norm": 0.6636178616387889, + "learning_rate": 4.373333333333334e-05, + "loss": 0.5681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5657886862754822, + "step": 165, + "valid_targets_mean": 28936.7, + "valid_targets_min": 22572 + }, + { + "epoch": 0.1814300960512273, + "grad_norm": 0.613994440314584, + "learning_rate": 4.5066666666666667e-05, + "loss": 0.5683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.56809401512146, + "step": 170, + "valid_targets_mean": 28852.1, + "valid_targets_min": 20877 + }, + { + "epoch": 0.18676627534685167, + "grad_norm": 0.6899646348527206, + "learning_rate": 4.64e-05, + "loss": 0.5667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5632824897766113, + "step": 175, + "valid_targets_mean": 28809.7, + "valid_targets_min": 21590 + }, + { + "epoch": 0.192102454642476, + "grad_norm": 0.4873629937543719, + "learning_rate": 4.773333333333333e-05, + "loss": 0.5643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.565109133720398, + "step": 180, + "valid_targets_mean": 28640.0, + "valid_targets_min": 17506 + }, + { + "epoch": 0.19743863393810032, + "grad_norm": 0.5777759535879822, + "learning_rate": 4.906666666666667e-05, + "loss": 0.5669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5687307119369507, + "step": 185, + "valid_targets_mean": 29064.5, + "valid_targets_min": 24174 + }, + { + "epoch": 0.20277481323372465, + "grad_norm": 0.6884108094887561, + "learning_rate": 5.0400000000000005e-05, + "loss": 0.5667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5649628043174744, + "step": 190, + "valid_targets_mean": 28915.0, + "valid_targets_min": 22022 + }, + { + "epoch": 0.20811099252934898, + "grad_norm": 0.6017305517883439, + "learning_rate": 5.1733333333333335e-05, + "loss": 0.5628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5632403492927551, + "step": 195, + "valid_targets_mean": 28751.2, + "valid_targets_min": 19909 + }, + { + "epoch": 0.21344717182497333, + "grad_norm": 0.6802658164600427, + "learning_rate": 5.3066666666666665e-05, + "loss": 0.5646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5695996284484863, + "step": 200, + "valid_targets_mean": 28698.4, + "valid_targets_min": 21957 + }, + { + "epoch": 0.21878335112059766, + "grad_norm": 0.4538292008860434, + "learning_rate": 5.440000000000001e-05, + "loss": 0.5618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5593487024307251, + "step": 205, + "valid_targets_mean": 28750.0, + "valid_targets_min": 17693 + }, + { + "epoch": 0.22411953041622198, + "grad_norm": 0.641577142782705, + "learning_rate": 5.573333333333334e-05, + "loss": 0.5645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5658643245697021, + "step": 210, + "valid_targets_mean": 28656.9, + "valid_targets_min": 17376 + }, + { + "epoch": 0.2294557097118463, + "grad_norm": 0.5170054885262663, + "learning_rate": 5.706666666666667e-05, + "loss": 0.5614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5651887655258179, + "step": 215, + "valid_targets_mean": 28857.5, + "valid_targets_min": 23359 + }, + { + "epoch": 0.23479188900747064, + "grad_norm": 0.6177322178483213, + "learning_rate": 5.8399999999999997e-05, + "loss": 0.5601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5584475994110107, + "step": 220, + "valid_targets_mean": 28901.5, + "valid_targets_min": 22575 + }, + { + "epoch": 0.240128068303095, + "grad_norm": 0.5026653012436633, + "learning_rate": 5.973333333333334e-05, + "loss": 0.5599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5588065385818481, + "step": 225, + "valid_targets_mean": 28699.3, + "valid_targets_min": 21859 + }, + { + "epoch": 0.24546424759871932, + "grad_norm": 0.4995892379669603, + "learning_rate": 6.106666666666667e-05, + "loss": 0.5607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5605010986328125, + "step": 230, + "valid_targets_mean": 28722.3, + "valid_targets_min": 22420 + }, + { + "epoch": 0.2508004268943437, + "grad_norm": 0.650437313972496, + "learning_rate": 6.24e-05, + "loss": 0.5587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5637643933296204, + "step": 235, + "valid_targets_mean": 28703.3, + "valid_targets_min": 21783 + }, + { + "epoch": 0.256136606189968, + "grad_norm": 0.4374501963289653, + "learning_rate": 6.373333333333333e-05, + "loss": 0.5571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.558539092540741, + "step": 240, + "valid_targets_mean": 28831.0, + "valid_targets_min": 19105 + }, + { + "epoch": 0.26147278548559233, + "grad_norm": 0.5120460721645188, + "learning_rate": 6.506666666666666e-05, + "loss": 0.5562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5536611080169678, + "step": 245, + "valid_targets_mean": 28847.3, + "valid_targets_min": 20476 + }, + { + "epoch": 0.26680896478121663, + "grad_norm": 0.6197628197231306, + "learning_rate": 6.64e-05, + "loss": 0.5557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5548427104949951, + "step": 250, + "valid_targets_mean": 28817.5, + "valid_targets_min": 19630 + }, + { + "epoch": 0.272145144076841, + "grad_norm": 0.71234145392978, + "learning_rate": 6.773333333333333e-05, + "loss": 0.5573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5544036626815796, + "step": 255, + "valid_targets_mean": 28778.9, + "valid_targets_min": 20199 + }, + { + "epoch": 0.27748132337246534, + "grad_norm": 0.6513418768796597, + "learning_rate": 6.906666666666667e-05, + "loss": 0.5536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5518766045570374, + "step": 260, + "valid_targets_mean": 28928.9, + "valid_targets_min": 20515 + }, + { + "epoch": 0.28281750266808964, + "grad_norm": 0.6264548727276208, + "learning_rate": 7.04e-05, + "loss": 0.5542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5519798994064331, + "step": 265, + "valid_targets_mean": 28822.4, + "valid_targets_min": 18953 + }, + { + "epoch": 0.288153681963714, + "grad_norm": 0.9786656061806465, + "learning_rate": 7.173333333333335e-05, + "loss": 0.5552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5583288669586182, + "step": 270, + "valid_targets_mean": 28753.7, + "valid_targets_min": 20559 + }, + { + "epoch": 0.2934898612593383, + "grad_norm": 0.49572263802680083, + "learning_rate": 7.306666666666668e-05, + "loss": 0.5525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5550556182861328, + "step": 275, + "valid_targets_mean": 28987.9, + "valid_targets_min": 19281 + }, + { + "epoch": 0.29882604055496265, + "grad_norm": 0.6362766358978091, + "learning_rate": 7.44e-05, + "loss": 0.5558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.549485445022583, + "step": 280, + "valid_targets_mean": 28889.4, + "valid_targets_min": 19873 + }, + { + "epoch": 0.304162219850587, + "grad_norm": 0.946560173004172, + "learning_rate": 7.573333333333334e-05, + "loss": 0.5542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5535620450973511, + "step": 285, + "valid_targets_mean": 29046.5, + "valid_targets_min": 23402 + }, + { + "epoch": 0.3094983991462113, + "grad_norm": 0.6213054757676977, + "learning_rate": 7.706666666666668e-05, + "loss": 0.5512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5472004413604736, + "step": 290, + "valid_targets_mean": 28608.9, + "valid_targets_min": 17886 + }, + { + "epoch": 0.31483457844183566, + "grad_norm": 0.8231565352486132, + "learning_rate": 7.840000000000001e-05, + "loss": 0.5513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5477820038795471, + "step": 295, + "valid_targets_mean": 28731.2, + "valid_targets_min": 21340 + }, + { + "epoch": 0.32017075773745995, + "grad_norm": 0.7927421222958255, + "learning_rate": 7.973333333333334e-05, + "loss": 0.5508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5506869554519653, + "step": 300, + "valid_targets_mean": 28976.3, + "valid_targets_min": 22880 + }, + { + "epoch": 0.3255069370330843, + "grad_norm": 0.7032608359190291, + "learning_rate": 8.106666666666667e-05, + "loss": 0.5524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5539842844009399, + "step": 305, + "valid_targets_mean": 28901.7, + "valid_targets_min": 21855 + }, + { + "epoch": 0.33084311632870866, + "grad_norm": 0.6731165496458789, + "learning_rate": 8.24e-05, + "loss": 0.5456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.538582444190979, + "step": 310, + "valid_targets_mean": 28647.1, + "valid_targets_min": 17737 + }, + { + "epoch": 0.33617929562433296, + "grad_norm": 0.7605105711198032, + "learning_rate": 8.373333333333334e-05, + "loss": 0.5485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5467817783355713, + "step": 315, + "valid_targets_mean": 28880.8, + "valid_targets_min": 19209 + }, + { + "epoch": 0.3415154749199573, + "grad_norm": 0.7701934663081149, + "learning_rate": 8.506666666666667e-05, + "loss": 0.5457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5447990894317627, + "step": 320, + "valid_targets_mean": 28782.5, + "valid_targets_min": 21917 + }, + { + "epoch": 0.3468516542155816, + "grad_norm": 0.5887577074410728, + "learning_rate": 8.64e-05, + "loss": 0.55, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5493910312652588, + "step": 325, + "valid_targets_mean": 28818.9, + "valid_targets_min": 20918 + }, + { + "epoch": 0.35218783351120597, + "grad_norm": 0.7582401178842324, + "learning_rate": 8.773333333333333e-05, + "loss": 0.546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5477738380432129, + "step": 330, + "valid_targets_mean": 28822.9, + "valid_targets_min": 21199 + }, + { + "epoch": 0.3575240128068303, + "grad_norm": 0.637643072735598, + "learning_rate": 8.906666666666667e-05, + "loss": 0.5464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.54499351978302, + "step": 335, + "valid_targets_mean": 28851.6, + "valid_targets_min": 20387 + }, + { + "epoch": 0.3628601921024546, + "grad_norm": 0.6724352862905262, + "learning_rate": 9.04e-05, + "loss": 0.5445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5406051874160767, + "step": 340, + "valid_targets_mean": 28807.4, + "valid_targets_min": 20681 + }, + { + "epoch": 0.368196371398079, + "grad_norm": 0.679305938726839, + "learning_rate": 9.173333333333333e-05, + "loss": 0.5452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5453157424926758, + "step": 345, + "valid_targets_mean": 28767.2, + "valid_targets_min": 17033 + }, + { + "epoch": 0.37353255069370334, + "grad_norm": 0.7960103561430129, + "learning_rate": 9.306666666666667e-05, + "loss": 0.5437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5461529493331909, + "step": 350, + "valid_targets_mean": 28861.1, + "valid_targets_min": 20183 + }, + { + "epoch": 0.37886872998932764, + "grad_norm": 0.7876674594635931, + "learning_rate": 9.44e-05, + "loss": 0.5431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5420316457748413, + "step": 355, + "valid_targets_mean": 28843.2, + "valid_targets_min": 18915 + }, + { + "epoch": 0.384204909284952, + "grad_norm": 0.5866144039786133, + "learning_rate": 9.573333333333335e-05, + "loss": 0.5407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5363604426383972, + "step": 360, + "valid_targets_mean": 28832.0, + "valid_targets_min": 22102 + }, + { + "epoch": 0.3895410885805763, + "grad_norm": 0.8109992871458446, + "learning_rate": 9.706666666666668e-05, + "loss": 0.5418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5428715944290161, + "step": 365, + "valid_targets_mean": 28704.1, + "valid_targets_min": 19957 + }, + { + "epoch": 0.39487726787620064, + "grad_norm": 0.8409633161647663, + "learning_rate": 9.84e-05, + "loss": 0.5414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5436188578605652, + "step": 370, + "valid_targets_mean": 28830.3, + "valid_targets_min": 21361 + }, + { + "epoch": 0.400213447171825, + "grad_norm": 0.6132368932425373, + "learning_rate": 9.973333333333334e-05, + "loss": 0.5422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5386438369750977, + "step": 375, + "valid_targets_mean": 28820.3, + "valid_targets_min": 22322 + }, + { + "epoch": 0.4055496264674493, + "grad_norm": 0.8058492330870486, + "learning_rate": 0.00010106666666666667, + "loss": 0.5416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.53936368227005, + "step": 380, + "valid_targets_mean": 28791.6, + "valid_targets_min": 21860 + }, + { + "epoch": 0.41088580576307365, + "grad_norm": 0.7700529206677325, + "learning_rate": 0.00010240000000000001, + "loss": 0.5404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5447088479995728, + "step": 385, + "valid_targets_mean": 28704.1, + "valid_targets_min": 20010 + }, + { + "epoch": 0.41622198505869795, + "grad_norm": 0.8893683706573925, + "learning_rate": 0.00010373333333333335, + "loss": 0.5397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5416263341903687, + "step": 390, + "valid_targets_mean": 28805.6, + "valid_targets_min": 20428 + }, + { + "epoch": 0.4215581643543223, + "grad_norm": 1.0117185257792445, + "learning_rate": 0.00010506666666666667, + "loss": 0.5425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5466989874839783, + "step": 395, + "valid_targets_mean": 28593.4, + "valid_targets_min": 20891 + }, + { + "epoch": 0.42689434364994666, + "grad_norm": 0.8925858685066481, + "learning_rate": 0.00010640000000000001, + "loss": 0.5411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5401574969291687, + "step": 400, + "valid_targets_mean": 29122.6, + "valid_targets_min": 19030 + }, + { + "epoch": 0.43223052294557096, + "grad_norm": 0.7349507561363565, + "learning_rate": 0.00010773333333333333, + "loss": 0.5399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5398991107940674, + "step": 405, + "valid_targets_mean": 28741.1, + "valid_targets_min": 21502 + }, + { + "epoch": 0.4375667022411953, + "grad_norm": 1.198184363542487, + "learning_rate": 0.00010906666666666667, + "loss": 0.5389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5372830033302307, + "step": 410, + "valid_targets_mean": 28847.6, + "valid_targets_min": 22709 + }, + { + "epoch": 0.4429028815368196, + "grad_norm": 0.9502266248869049, + "learning_rate": 0.00011040000000000001, + "loss": 0.5394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5405418276786804, + "step": 415, + "valid_targets_mean": 28784.6, + "valid_targets_min": 19776 + }, + { + "epoch": 0.44823906083244397, + "grad_norm": 0.7936068497230154, + "learning_rate": 0.00011173333333333333, + "loss": 0.539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5384629964828491, + "step": 420, + "valid_targets_mean": 28831.8, + "valid_targets_min": 23168 + }, + { + "epoch": 0.4535752401280683, + "grad_norm": 0.7204539211375129, + "learning_rate": 0.00011306666666666667, + "loss": 0.5382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5382857322692871, + "step": 425, + "valid_targets_mean": 28729.3, + "valid_targets_min": 18829 + }, + { + "epoch": 0.4589114194236926, + "grad_norm": 0.9715280954871497, + "learning_rate": 0.0001144, + "loss": 0.5345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5296086072921753, + "step": 430, + "valid_targets_mean": 28995.2, + "valid_targets_min": 21516 + }, + { + "epoch": 0.464247598719317, + "grad_norm": 1.1643790183048932, + "learning_rate": 0.00011573333333333333, + "loss": 0.5367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.537731409072876, + "step": 435, + "valid_targets_mean": 28795.0, + "valid_targets_min": 19489 + }, + { + "epoch": 0.4695837780149413, + "grad_norm": 0.8770740878776596, + "learning_rate": 0.00011706666666666668, + "loss": 0.5384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5397451519966125, + "step": 440, + "valid_targets_mean": 28636.4, + "valid_targets_min": 16244 + }, + { + "epoch": 0.47491995731056563, + "grad_norm": 0.9344853601422126, + "learning_rate": 0.0001184, + "loss": 0.5379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5392533540725708, + "step": 445, + "valid_targets_mean": 28694.7, + "valid_targets_min": 22274 + }, + { + "epoch": 0.48025613660619, + "grad_norm": 0.7668304201033254, + "learning_rate": 0.00011973333333333335, + "loss": 0.5365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5335771441459656, + "step": 450, + "valid_targets_mean": 28841.2, + "valid_targets_min": 21563 + }, + { + "epoch": 0.4855923159018143, + "grad_norm": 0.6698987095508048, + "learning_rate": 0.00012106666666666666, + "loss": 0.5357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.534684956073761, + "step": 455, + "valid_targets_mean": 28906.0, + "valid_targets_min": 21832 + }, + { + "epoch": 0.49092849519743864, + "grad_norm": 0.9076765335380783, + "learning_rate": 0.0001224, + "loss": 0.5338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5326440334320068, + "step": 460, + "valid_targets_mean": 28854.4, + "valid_targets_min": 22257 + }, + { + "epoch": 0.49626467449306294, + "grad_norm": 0.8366795432780757, + "learning_rate": 0.00012373333333333335, + "loss": 0.5304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5290672183036804, + "step": 465, + "valid_targets_mean": 28736.9, + "valid_targets_min": 18357 + }, + { + "epoch": 0.5016008537886874, + "grad_norm": 1.0802707635836115, + "learning_rate": 0.00012506666666666665, + "loss": 0.5353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5346341133117676, + "step": 470, + "valid_targets_mean": 28902.0, + "valid_targets_min": 22967 + }, + { + "epoch": 0.5069370330843116, + "grad_norm": 0.9609279719889722, + "learning_rate": 0.0001264, + "loss": 0.5326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5342420935630798, + "step": 475, + "valid_targets_mean": 28665.1, + "valid_targets_min": 21412 + }, + { + "epoch": 0.512273212379936, + "grad_norm": 1.3158372695521554, + "learning_rate": 0.00012773333333333334, + "loss": 0.5344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5368781089782715, + "step": 480, + "valid_targets_mean": 28766.8, + "valid_targets_min": 19331 + }, + { + "epoch": 0.5176093916755603, + "grad_norm": 0.9795998874478892, + "learning_rate": 0.00012906666666666667, + "loss": 0.5342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5322054624557495, + "step": 485, + "valid_targets_mean": 28829.6, + "valid_targets_min": 20404 + }, + { + "epoch": 0.5229455709711847, + "grad_norm": 0.6778702038700283, + "learning_rate": 0.0001304, + "loss": 0.5334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5259044170379639, + "step": 490, + "valid_targets_mean": 28834.2, + "valid_targets_min": 22315 + }, + { + "epoch": 0.528281750266809, + "grad_norm": 1.317650223882812, + "learning_rate": 0.00013173333333333333, + "loss": 0.5318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5341742038726807, + "step": 495, + "valid_targets_mean": 28683.8, + "valid_targets_min": 19502 + }, + { + "epoch": 0.5336179295624333, + "grad_norm": 1.0156689995894708, + "learning_rate": 0.00013306666666666668, + "loss": 0.535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.535778284072876, + "step": 500, + "valid_targets_mean": 28661.4, + "valid_targets_min": 20971 + }, + { + "epoch": 0.5389541088580576, + "grad_norm": 1.0878575161019286, + "learning_rate": 0.00013440000000000001, + "loss": 0.5318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5342371463775635, + "step": 505, + "valid_targets_mean": 28679.2, + "valid_targets_min": 18819 + }, + { + "epoch": 0.544290288153682, + "grad_norm": 1.0272627391559297, + "learning_rate": 0.00013573333333333334, + "loss": 0.5293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5333046913146973, + "step": 510, + "valid_targets_mean": 28750.6, + "valid_targets_min": 21745 + }, + { + "epoch": 0.5496264674493063, + "grad_norm": 0.8126962382300175, + "learning_rate": 0.00013706666666666667, + "loss": 0.5288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5308830738067627, + "step": 515, + "valid_targets_mean": 28827.4, + "valid_targets_min": 19230 + }, + { + "epoch": 0.5549626467449307, + "grad_norm": 0.7393375684190655, + "learning_rate": 0.0001384, + "loss": 0.5318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5321455597877502, + "step": 520, + "valid_targets_mean": 28860.7, + "valid_targets_min": 20761 + }, + { + "epoch": 0.5602988260405549, + "grad_norm": 0.791448603346915, + "learning_rate": 0.00013973333333333333, + "loss": 0.5311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5327831506729126, + "step": 525, + "valid_targets_mean": 28866.7, + "valid_targets_min": 21514 + }, + { + "epoch": 0.5656350053361793, + "grad_norm": 0.7635180566237668, + "learning_rate": 0.0001410666666666667, + "loss": 0.5292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5294798016548157, + "step": 530, + "valid_targets_mean": 28815.9, + "valid_targets_min": 22727 + }, + { + "epoch": 0.5709711846318036, + "grad_norm": 0.7299144969095928, + "learning_rate": 0.0001424, + "loss": 0.5276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5238897800445557, + "step": 535, + "valid_targets_mean": 28845.8, + "valid_targets_min": 22600 + }, + { + "epoch": 0.576307363927428, + "grad_norm": 0.9463246724807779, + "learning_rate": 0.00014373333333333335, + "loss": 0.5289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5301527976989746, + "step": 540, + "valid_targets_mean": 28835.9, + "valid_targets_min": 22871 + }, + { + "epoch": 0.5816435432230523, + "grad_norm": 0.7734584962993285, + "learning_rate": 0.00014506666666666668, + "loss": 0.5251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5256263017654419, + "step": 545, + "valid_targets_mean": 28784.2, + "valid_targets_min": 22150 + }, + { + "epoch": 0.5869797225186766, + "grad_norm": 1.2225580809817047, + "learning_rate": 0.0001464, + "loss": 0.5298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5331275463104248, + "step": 550, + "valid_targets_mean": 28789.1, + "valid_targets_min": 22110 + }, + { + "epoch": 0.5923159018143009, + "grad_norm": 0.7567944602442186, + "learning_rate": 0.00014773333333333334, + "loss": 0.5287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5319618582725525, + "step": 555, + "valid_targets_mean": 28775.0, + "valid_targets_min": 21938 + }, + { + "epoch": 0.5976520811099253, + "grad_norm": 0.9601392530255699, + "learning_rate": 0.00014906666666666667, + "loss": 0.5278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5328255295753479, + "step": 560, + "valid_targets_mean": 28883.7, + "valid_targets_min": 16759 + }, + { + "epoch": 0.6029882604055496, + "grad_norm": 1.1840261155964624, + "learning_rate": 0.0001504, + "loss": 0.5264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5287203788757324, + "step": 565, + "valid_targets_mean": 28914.4, + "valid_targets_min": 23391 + }, + { + "epoch": 0.608324439701174, + "grad_norm": 0.7959310356438406, + "learning_rate": 0.00015173333333333335, + "loss": 0.5298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5267726182937622, + "step": 570, + "valid_targets_mean": 28953.4, + "valid_targets_min": 22290 + }, + { + "epoch": 0.6136606189967982, + "grad_norm": 0.8250892679360178, + "learning_rate": 0.00015306666666666666, + "loss": 0.5272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5269379019737244, + "step": 575, + "valid_targets_mean": 28862.8, + "valid_targets_min": 20440 + }, + { + "epoch": 0.6189967982924226, + "grad_norm": 1.1422164796143262, + "learning_rate": 0.0001544, + "loss": 0.5277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5261090993881226, + "step": 580, + "valid_targets_mean": 28641.9, + "valid_targets_min": 13903 + }, + { + "epoch": 0.624332977588047, + "grad_norm": 0.6963629627094391, + "learning_rate": 0.00015573333333333334, + "loss": 0.5259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5219688415527344, + "step": 585, + "valid_targets_mean": 29003.0, + "valid_targets_min": 20872 + }, + { + "epoch": 0.6296691568836713, + "grad_norm": 0.9603755444061319, + "learning_rate": 0.00015706666666666667, + "loss": 0.5245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5274404287338257, + "step": 590, + "valid_targets_mean": 29018.0, + "valid_targets_min": 18214 + }, + { + "epoch": 0.6350053361792957, + "grad_norm": 0.9742014368447384, + "learning_rate": 0.00015840000000000003, + "loss": 0.5261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5234408378601074, + "step": 595, + "valid_targets_mean": 28871.5, + "valid_targets_min": 21415 + }, + { + "epoch": 0.6403415154749199, + "grad_norm": 1.1012816833594785, + "learning_rate": 0.00015973333333333333, + "loss": 0.5242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5250797271728516, + "step": 600, + "valid_targets_mean": 28953.2, + "valid_targets_min": 23044 + }, + { + "epoch": 0.6456776947705443, + "grad_norm": 1.17331547694115, + "learning_rate": 0.0001610666666666667, + "loss": 0.5263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5289738774299622, + "step": 605, + "valid_targets_mean": 29002.3, + "valid_targets_min": 23780 + }, + { + "epoch": 0.6510138740661686, + "grad_norm": 0.7971149521840055, + "learning_rate": 0.00016240000000000002, + "loss": 0.5238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5232728719711304, + "step": 610, + "valid_targets_mean": 28782.3, + "valid_targets_min": 21471 + }, + { + "epoch": 0.656350053361793, + "grad_norm": 0.8398684106900386, + "learning_rate": 0.00016373333333333335, + "loss": 0.5227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.525774359703064, + "step": 615, + "valid_targets_mean": 28889.3, + "valid_targets_min": 19959 + }, + { + "epoch": 0.6616862326574173, + "grad_norm": 0.7198912861291683, + "learning_rate": 0.00016506666666666668, + "loss": 0.5244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5229619741439819, + "step": 620, + "valid_targets_mean": 28760.8, + "valid_targets_min": 17978 + }, + { + "epoch": 0.6670224119530416, + "grad_norm": 0.9190832295110807, + "learning_rate": 0.0001664, + "loss": 0.5231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.52231764793396, + "step": 625, + "valid_targets_mean": 28903.8, + "valid_targets_min": 18702 + }, + { + "epoch": 0.6723585912486659, + "grad_norm": 0.9169151703365269, + "learning_rate": 0.00016773333333333334, + "loss": 0.5243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5236749649047852, + "step": 630, + "valid_targets_mean": 28919.6, + "valid_targets_min": 18070 + }, + { + "epoch": 0.6776947705442903, + "grad_norm": 0.9440014677930404, + "learning_rate": 0.0001690666666666667, + "loss": 0.5219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.519328236579895, + "step": 635, + "valid_targets_mean": 28835.2, + "valid_targets_min": 22006 + }, + { + "epoch": 0.6830309498399146, + "grad_norm": 1.1226009385172517, + "learning_rate": 0.0001704, + "loss": 0.5226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.524982213973999, + "step": 640, + "valid_targets_mean": 28876.1, + "valid_targets_min": 17173 + }, + { + "epoch": 0.688367129135539, + "grad_norm": 0.7893843871981923, + "learning_rate": 0.00017173333333333335, + "loss": 0.5218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5177816152572632, + "step": 645, + "valid_targets_mean": 28941.9, + "valid_targets_min": 23033 + }, + { + "epoch": 0.6937033084311632, + "grad_norm": 0.9900573740334639, + "learning_rate": 0.00017306666666666665, + "loss": 0.5232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5200702548027039, + "step": 650, + "valid_targets_mean": 28843.3, + "valid_targets_min": 22189 + }, + { + "epoch": 0.6990394877267876, + "grad_norm": 1.050873220431905, + "learning_rate": 0.0001744, + "loss": 0.5218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.521094799041748, + "step": 655, + "valid_targets_mean": 28854.3, + "valid_targets_min": 22418 + }, + { + "epoch": 0.7043756670224119, + "grad_norm": 0.9507076490054547, + "learning_rate": 0.00017573333333333334, + "loss": 0.521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5252164006233215, + "step": 660, + "valid_targets_mean": 28942.5, + "valid_targets_min": 21262 + }, + { + "epoch": 0.7097118463180363, + "grad_norm": 0.9417609906453398, + "learning_rate": 0.00017706666666666667, + "loss": 0.5215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5187242031097412, + "step": 665, + "valid_targets_mean": 28845.2, + "valid_targets_min": 19004 + }, + { + "epoch": 0.7150480256136607, + "grad_norm": 0.7494082136939058, + "learning_rate": 0.0001784, + "loss": 0.5203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5231742858886719, + "step": 670, + "valid_targets_mean": 28843.6, + "valid_targets_min": 19456 + }, + { + "epoch": 0.7203842049092849, + "grad_norm": 1.0576258309034712, + "learning_rate": 0.00017973333333333333, + "loss": 0.5228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5267496109008789, + "step": 675, + "valid_targets_mean": 28791.6, + "valid_targets_min": 18590 + }, + { + "epoch": 0.7257203842049093, + "grad_norm": 1.0947745640355913, + "learning_rate": 0.00018106666666666669, + "loss": 0.5218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5197322368621826, + "step": 680, + "valid_targets_mean": 28867.4, + "valid_targets_min": 23073 + }, + { + "epoch": 0.7310565635005336, + "grad_norm": 0.8804688508632426, + "learning_rate": 0.00018240000000000002, + "loss": 0.5181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5168286561965942, + "step": 685, + "valid_targets_mean": 28865.1, + "valid_targets_min": 21222 + }, + { + "epoch": 0.736392742796158, + "grad_norm": 0.8377434122835627, + "learning_rate": 0.00018373333333333335, + "loss": 0.5199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5161206126213074, + "step": 690, + "valid_targets_mean": 28880.2, + "valid_targets_min": 19297 + }, + { + "epoch": 0.7417289220917823, + "grad_norm": 1.130069939334258, + "learning_rate": 0.00018506666666666667, + "loss": 0.5186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5214225053787231, + "step": 695, + "valid_targets_mean": 28741.1, + "valid_targets_min": 19445 + }, + { + "epoch": 0.7470651013874067, + "grad_norm": 0.8103866371897044, + "learning_rate": 0.00018640000000000003, + "loss": 0.5176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5210945010185242, + "step": 700, + "valid_targets_mean": 28913.4, + "valid_targets_min": 22165 + }, + { + "epoch": 0.7524012806830309, + "grad_norm": 0.9569292836049637, + "learning_rate": 0.00018773333333333333, + "loss": 0.5171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5173620581626892, + "step": 705, + "valid_targets_mean": 28761.6, + "valid_targets_min": 16988 + }, + { + "epoch": 0.7577374599786553, + "grad_norm": 1.0348041744035754, + "learning_rate": 0.0001890666666666667, + "loss": 0.5182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5234087109565735, + "step": 710, + "valid_targets_mean": 28747.3, + "valid_targets_min": 17862 + }, + { + "epoch": 0.7630736392742796, + "grad_norm": 0.6395456940178995, + "learning_rate": 0.0001904, + "loss": 0.5134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5095891952514648, + "step": 715, + "valid_targets_mean": 28773.6, + "valid_targets_min": 19732 + }, + { + "epoch": 0.768409818569904, + "grad_norm": 0.9053417720870189, + "learning_rate": 0.00019173333333333335, + "loss": 0.5173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.516966700553894, + "step": 720, + "valid_targets_mean": 28985.3, + "valid_targets_min": 21313 + }, + { + "epoch": 0.7737459978655283, + "grad_norm": 1.1812927567242297, + "learning_rate": 0.00019306666666666668, + "loss": 0.516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5194029808044434, + "step": 725, + "valid_targets_mean": 28903.3, + "valid_targets_min": 18221 + }, + { + "epoch": 0.7790821771611526, + "grad_norm": 1.1415316033642284, + "learning_rate": 0.0001944, + "loss": 0.5167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5210375785827637, + "step": 730, + "valid_targets_mean": 28849.7, + "valid_targets_min": 20021 + }, + { + "epoch": 0.7844183564567769, + "grad_norm": 0.9275823164638465, + "learning_rate": 0.00019573333333333334, + "loss": 0.5156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5161601305007935, + "step": 735, + "valid_targets_mean": 28839.5, + "valid_targets_min": 21489 + }, + { + "epoch": 0.7897545357524013, + "grad_norm": 1.0879808817539458, + "learning_rate": 0.00019706666666666667, + "loss": 0.5153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5184020400047302, + "step": 740, + "valid_targets_mean": 28919.2, + "valid_targets_min": 21215 + }, + { + "epoch": 0.7950907150480256, + "grad_norm": 0.9695487034080649, + "learning_rate": 0.0001984, + "loss": 0.514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5139423608779907, + "step": 745, + "valid_targets_mean": 28771.1, + "valid_targets_min": 21422 + }, + { + "epoch": 0.80042689434365, + "grad_norm": 0.7526157185372891, + "learning_rate": 0.00019973333333333335, + "loss": 0.5153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5156712532043457, + "step": 750, + "valid_targets_mean": 28994.6, + "valid_targets_min": 22465 + }, + { + "epoch": 0.8057630736392742, + "grad_norm": 0.8534620058988129, + "learning_rate": 0.00019999982650101714, + "loss": 0.5113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5072367787361145, + "step": 755, + "valid_targets_mean": 28920.0, + "valid_targets_min": 18753 + }, + { + "epoch": 0.8110992529348986, + "grad_norm": 0.9207936626555079, + "learning_rate": 0.00019999912166243107, + "loss": 0.5131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.512698769569397, + "step": 760, + "valid_targets_mean": 29037.0, + "valid_targets_min": 19954 + }, + { + "epoch": 0.816435432230523, + "grad_norm": 0.5737733806118188, + "learning_rate": 0.00019999787464437395, + "loss": 0.5116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5104231238365173, + "step": 765, + "valid_targets_mean": 28870.6, + "valid_targets_min": 18563 + }, + { + "epoch": 0.8217716115261473, + "grad_norm": 0.9923597897717296, + "learning_rate": 0.0001999960854536069, + "loss": 0.5189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5126428604125977, + "step": 770, + "valid_targets_mean": 28757.4, + "valid_targets_min": 22298 + }, + { + "epoch": 0.8271077908217717, + "grad_norm": 0.9603156077658574, + "learning_rate": 0.00019999375409983057, + "loss": 0.5155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5126303434371948, + "step": 775, + "valid_targets_mean": 28917.4, + "valid_targets_min": 21333 + }, + { + "epoch": 0.8324439701173959, + "grad_norm": 0.8981896229393682, + "learning_rate": 0.00019999088059568536, + "loss": 0.5129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5156265497207642, + "step": 780, + "valid_targets_mean": 28922.4, + "valid_targets_min": 23138 + }, + { + "epoch": 0.8377801494130203, + "grad_norm": 0.7400545260515701, + "learning_rate": 0.0001999874649567508, + "loss": 0.5128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5122023820877075, + "step": 785, + "valid_targets_mean": 28929.7, + "valid_targets_min": 22171 + }, + { + "epoch": 0.8431163287086446, + "grad_norm": 0.8913220008247429, + "learning_rate": 0.00019998350720154605, + "loss": 0.513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5137268304824829, + "step": 790, + "valid_targets_mean": 28965.1, + "valid_targets_min": 19535 + }, + { + "epoch": 0.848452508004269, + "grad_norm": 0.6618361104222048, + "learning_rate": 0.00019997900735152936, + "loss": 0.5111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5073462724685669, + "step": 795, + "valid_targets_mean": 28941.9, + "valid_targets_min": 20187 + }, + { + "epoch": 0.8537886872998933, + "grad_norm": 0.9835517191921062, + "learning_rate": 0.0001999739654310982, + "loss": 0.514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5118036270141602, + "step": 800, + "valid_targets_mean": 28857.4, + "valid_targets_min": 20301 + }, + { + "epoch": 0.8591248665955176, + "grad_norm": 0.7167219489963145, + "learning_rate": 0.00019996838146758915, + "loss": 0.5115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.513766884803772, + "step": 805, + "valid_targets_mean": 28856.1, + "valid_targets_min": 19253 + }, + { + "epoch": 0.8644610458911419, + "grad_norm": 0.8234050116161585, + "learning_rate": 0.00019996225549127753, + "loss": 0.5119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5139243006706238, + "step": 810, + "valid_targets_mean": 28901.4, + "valid_targets_min": 21836 + }, + { + "epoch": 0.8697972251867663, + "grad_norm": 1.0887508688343202, + "learning_rate": 0.0001999555875353774, + "loss": 0.5111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5112956762313843, + "step": 815, + "valid_targets_mean": 28909.4, + "valid_targets_min": 20239 + }, + { + "epoch": 0.8751334044823906, + "grad_norm": 0.876878433615169, + "learning_rate": 0.0001999483776360414, + "loss": 0.5101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5132863521575928, + "step": 820, + "valid_targets_mean": 28695.7, + "valid_targets_min": 18285 + }, + { + "epoch": 0.880469583778015, + "grad_norm": 0.8413466528813544, + "learning_rate": 0.00019994062583236045, + "loss": 0.5078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5094743967056274, + "step": 825, + "valid_targets_mean": 28791.7, + "valid_targets_min": 20225 + }, + { + "epoch": 0.8858057630736392, + "grad_norm": 0.8600849420247135, + "learning_rate": 0.00019993233216636358, + "loss": 0.5115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5101096630096436, + "step": 830, + "valid_targets_mean": 28825.4, + "valid_targets_min": 22899 + }, + { + "epoch": 0.8911419423692636, + "grad_norm": 0.6651167236591127, + "learning_rate": 0.0001999234966830178, + "loss": 0.5083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5113962888717651, + "step": 835, + "valid_targets_mean": 28849.6, + "valid_targets_min": 21756 + }, + { + "epoch": 0.8964781216648879, + "grad_norm": 1.0818540858113659, + "learning_rate": 0.00019991411943022768, + "loss": 0.5114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5127806067466736, + "step": 840, + "valid_targets_mean": 29012.6, + "valid_targets_min": 20124 + }, + { + "epoch": 0.9018143009605123, + "grad_norm": 0.7661149143265992, + "learning_rate": 0.00019990420045883518, + "loss": 0.5093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5133446455001831, + "step": 845, + "valid_targets_mean": 28692.7, + "valid_targets_min": 21532 + }, + { + "epoch": 0.9071504802561366, + "grad_norm": 0.7850553702573292, + "learning_rate": 0.00019989373982261946, + "loss": 0.5075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.508018970489502, + "step": 850, + "valid_targets_mean": 28888.1, + "valid_targets_min": 22096 + }, + { + "epoch": 0.9124866595517609, + "grad_norm": 0.957979899606732, + "learning_rate": 0.0001998827375782964, + "loss": 0.508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5046581029891968, + "step": 855, + "valid_targets_mean": 28860.0, + "valid_targets_min": 18914 + }, + { + "epoch": 0.9178228388473852, + "grad_norm": 0.8046053319649763, + "learning_rate": 0.0001998711937855184, + "loss": 0.5064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5036020278930664, + "step": 860, + "valid_targets_mean": 28765.5, + "valid_targets_min": 17484 + }, + { + "epoch": 0.9231590181430096, + "grad_norm": 0.7647829514060208, + "learning_rate": 0.00019985910850687413, + "loss": 0.5087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5098488330841064, + "step": 865, + "valid_targets_mean": 28809.4, + "valid_targets_min": 21450 + }, + { + "epoch": 0.928495197438634, + "grad_norm": 0.8215795314493268, + "learning_rate": 0.00019984648180788804, + "loss": 0.5043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5045159459114075, + "step": 870, + "valid_targets_mean": 28818.4, + "valid_targets_min": 19185 + }, + { + "epoch": 0.9338313767342583, + "grad_norm": 0.8191498775289832, + "learning_rate": 0.00019983331375702016, + "loss": 0.5032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5012009143829346, + "step": 875, + "valid_targets_mean": 28765.6, + "valid_targets_min": 18225 + }, + { + "epoch": 0.9391675560298826, + "grad_norm": 0.6547442901261715, + "learning_rate": 0.00019981960442566553, + "loss": 0.5054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5023826360702515, + "step": 880, + "valid_targets_mean": 28933.4, + "valid_targets_min": 20988 + }, + { + "epoch": 0.9445037353255069, + "grad_norm": 0.7730738931584743, + "learning_rate": 0.000199805353888154, + "loss": 0.5052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5060917139053345, + "step": 885, + "valid_targets_mean": 28948.6, + "valid_targets_min": 18118 + }, + { + "epoch": 0.9498399146211313, + "grad_norm": 0.7811823083404794, + "learning_rate": 0.00019979056222174972, + "loss": 0.5058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5014211535453796, + "step": 890, + "valid_targets_mean": 28850.7, + "valid_targets_min": 22683 + }, + { + "epoch": 0.9551760939167556, + "grad_norm": 0.8319739955826252, + "learning_rate": 0.0001997752295066508, + "loss": 0.5051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5034403800964355, + "step": 895, + "valid_targets_mean": 29030.9, + "valid_targets_min": 19158 + }, + { + "epoch": 0.96051227321238, + "grad_norm": 0.7282233904577727, + "learning_rate": 0.0001997593558259888, + "loss": 0.5037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5004850625991821, + "step": 900, + "valid_targets_mean": 28900.0, + "valid_targets_min": 21973 + }, + { + "epoch": 0.9658484525080042, + "grad_norm": 1.3121763739243963, + "learning_rate": 0.00019974294126582826, + "loss": 0.5068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5165523886680603, + "step": 905, + "valid_targets_mean": 28758.2, + "valid_targets_min": 16583 + }, + { + "epoch": 0.9711846318036286, + "grad_norm": 1.0382108685703912, + "learning_rate": 0.00019972598591516642, + "loss": 0.5069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5057257413864136, + "step": 910, + "valid_targets_mean": 28712.1, + "valid_targets_min": 18488 + }, + { + "epoch": 0.9765208110992529, + "grad_norm": 0.7243104931637121, + "learning_rate": 0.00019970848986593248, + "loss": 0.5072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5118848085403442, + "step": 915, + "valid_targets_mean": 28952.7, + "valid_targets_min": 23291 + }, + { + "epoch": 0.9818569903948773, + "grad_norm": 0.7418714866265343, + "learning_rate": 0.00019969045321298725, + "loss": 0.5072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.506354033946991, + "step": 920, + "valid_targets_mean": 28869.9, + "valid_targets_min": 22004 + }, + { + "epoch": 0.9871931696905016, + "grad_norm": 0.7193766855257959, + "learning_rate": 0.00019967187605412262, + "loss": 0.5035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5024032592773438, + "step": 925, + "valid_targets_mean": 29032.0, + "valid_targets_min": 20042 + }, + { + "epoch": 0.9925293489861259, + "grad_norm": 0.8126562988133832, + "learning_rate": 0.00019965275849006102, + "loss": 0.5036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5021684765815735, + "step": 930, + "valid_targets_mean": 29001.4, + "valid_targets_min": 19151 + }, + { + "epoch": 0.9978655282817502, + "grad_norm": 0.729906175244711, + "learning_rate": 0.00019963310062445487, + "loss": 0.5022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.504224419593811, + "step": 935, + "valid_targets_mean": 28850.6, + "valid_targets_min": 22676 + }, + { + "epoch": 1.0032017075773747, + "grad_norm": 0.8857335048779751, + "learning_rate": 0.00019961290256388604, + "loss": 0.501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5056629776954651, + "step": 940, + "valid_targets_mean": 28686.6, + "valid_targets_min": 19870 + }, + { + "epoch": 1.0085378868729988, + "grad_norm": 0.7714220146823638, + "learning_rate": 0.00019959216441786524, + "loss": 0.4982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49670982360839844, + "step": 945, + "valid_targets_mean": 28700.9, + "valid_targets_min": 21769 + }, + { + "epoch": 1.0138740661686232, + "grad_norm": 0.8056427198462525, + "learning_rate": 0.00019957088629883135, + "loss": 0.4987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4962517023086548, + "step": 950, + "valid_targets_mean": 28952.8, + "valid_targets_min": 21902 + }, + { + "epoch": 1.0192102454642475, + "grad_norm": 0.5467811180272533, + "learning_rate": 0.00019954906832215103, + "loss": 0.4972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49847257137298584, + "step": 955, + "valid_targets_mean": 28823.4, + "valid_targets_min": 22059 + }, + { + "epoch": 1.024546424759872, + "grad_norm": 1.1696041007065114, + "learning_rate": 0.00019952671060611792, + "loss": 0.5008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5013498663902283, + "step": 960, + "valid_targets_mean": 28771.9, + "valid_targets_min": 19840 + }, + { + "epoch": 1.0298826040554963, + "grad_norm": 0.7944049884311958, + "learning_rate": 0.000199503813271952, + "loss": 0.5004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5025086402893066, + "step": 965, + "valid_targets_mean": 28858.1, + "valid_targets_min": 21155 + }, + { + "epoch": 1.0352187833511206, + "grad_norm": 0.6984525409939394, + "learning_rate": 0.00019948037644379905, + "loss": 0.4979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4993964433670044, + "step": 970, + "valid_targets_mean": 29067.5, + "valid_targets_min": 20386 + }, + { + "epoch": 1.040554962646745, + "grad_norm": 0.7610988833368267, + "learning_rate": 0.0001994564002487298, + "loss": 0.4993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5011273622512817, + "step": 975, + "valid_targets_mean": 28738.2, + "valid_targets_min": 19725 + }, + { + "epoch": 1.0458911419423693, + "grad_norm": 0.8782724630795262, + "learning_rate": 0.00019943188481673946, + "loss": 0.496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49535125494003296, + "step": 980, + "valid_targets_mean": 28957.1, + "valid_targets_min": 21418 + }, + { + "epoch": 1.0512273212379937, + "grad_norm": 0.8026106911230974, + "learning_rate": 0.00019940683028074675, + "loss": 0.4974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49761420488357544, + "step": 985, + "valid_targets_mean": 28961.3, + "valid_targets_min": 22587 + }, + { + "epoch": 1.056563500533618, + "grad_norm": 0.7959574826279312, + "learning_rate": 0.00019938123677659352, + "loss": 0.4973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49720531702041626, + "step": 990, + "valid_targets_mean": 29085.9, + "valid_targets_min": 23594 + }, + { + "epoch": 1.0618996798292422, + "grad_norm": 0.6285283346775284, + "learning_rate": 0.00019935510444304364, + "loss": 0.497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4930610656738281, + "step": 995, + "valid_targets_mean": 28967.6, + "valid_targets_min": 20145 + }, + { + "epoch": 1.0672358591248665, + "grad_norm": 0.9256561165916987, + "learning_rate": 0.00019932843342178254, + "loss": 0.4966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49736061692237854, + "step": 1000, + "valid_targets_mean": 29112.2, + "valid_targets_min": 23359 + }, + { + "epoch": 1.0725720384204909, + "grad_norm": 0.7313284825114021, + "learning_rate": 0.00019930122385741625, + "loss": 0.5005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5027332305908203, + "step": 1005, + "valid_targets_mean": 28745.6, + "valid_targets_min": 23079 + }, + { + "epoch": 1.0779082177161152, + "grad_norm": 0.7342069358731735, + "learning_rate": 0.00019927347589747082, + "loss": 0.4981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49375826120376587, + "step": 1010, + "valid_targets_mean": 29010.8, + "valid_targets_min": 19817 + }, + { + "epoch": 1.0832443970117396, + "grad_norm": 0.7672321385196467, + "learning_rate": 0.0001992451896923912, + "loss": 0.4957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49502408504486084, + "step": 1015, + "valid_targets_mean": 28889.4, + "valid_targets_min": 22578 + }, + { + "epoch": 1.088580576307364, + "grad_norm": 0.8271950543331492, + "learning_rate": 0.00019921636539554084, + "loss": 0.4977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4977359175682068, + "step": 1020, + "valid_targets_mean": 28720.5, + "valid_targets_min": 20101 + }, + { + "epoch": 1.0939167556029883, + "grad_norm": 0.7373693958339582, + "learning_rate": 0.00019918700316320044, + "loss": 0.4958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4936048686504364, + "step": 1025, + "valid_targets_mean": 29039.2, + "valid_targets_min": 21531 + }, + { + "epoch": 1.0992529348986126, + "grad_norm": 0.6151459246691149, + "learning_rate": 0.00019915710315456748, + "loss": 0.4927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4901975095272064, + "step": 1030, + "valid_targets_mean": 28976.8, + "valid_targets_min": 21388 + }, + { + "epoch": 1.104589114194237, + "grad_norm": 0.8842826137042493, + "learning_rate": 0.00019912666553175502, + "loss": 0.4944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4999387860298157, + "step": 1035, + "valid_targets_mean": 28779.4, + "valid_targets_min": 20977 + }, + { + "epoch": 1.1099252934898614, + "grad_norm": 0.809143508702625, + "learning_rate": 0.0001990956904597911, + "loss": 0.4937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48766613006591797, + "step": 1040, + "valid_targets_mean": 28873.3, + "valid_targets_min": 20283 + }, + { + "epoch": 1.1152614727854857, + "grad_norm": 0.8399523521804738, + "learning_rate": 0.0001990641781066177, + "loss": 0.4941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49525129795074463, + "step": 1045, + "valid_targets_mean": 28644.3, + "valid_targets_min": 18907 + }, + { + "epoch": 1.1205976520811098, + "grad_norm": 0.767585575338914, + "learning_rate": 0.00019903212864308975, + "loss": 0.4937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4969574809074402, + "step": 1050, + "valid_targets_mean": 28797.3, + "valid_targets_min": 21043 + }, + { + "epoch": 1.1259338313767342, + "grad_norm": 0.7413224378945623, + "learning_rate": 0.00019899954224297442, + "loss": 0.4953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49663180112838745, + "step": 1055, + "valid_targets_mean": 28732.7, + "valid_targets_min": 23116 + }, + { + "epoch": 1.1312700106723586, + "grad_norm": 0.7658592966316667, + "learning_rate": 0.00019896641908295006, + "loss": 0.4917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4922265410423279, + "step": 1060, + "valid_targets_mean": 28798.4, + "valid_targets_min": 20032 + }, + { + "epoch": 1.136606189967983, + "grad_norm": 0.6444301983737417, + "learning_rate": 0.00019893275934260523, + "loss": 0.4943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49562233686447144, + "step": 1065, + "valid_targets_mean": 28789.7, + "valid_targets_min": 21020 + }, + { + "epoch": 1.1419423692636073, + "grad_norm": 0.9172448797269533, + "learning_rate": 0.00019889856320443768, + "loss": 0.4932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4929046928882599, + "step": 1070, + "valid_targets_mean": 28764.4, + "valid_targets_min": 19307 + }, + { + "epoch": 1.1472785485592316, + "grad_norm": 0.726566080960612, + "learning_rate": 0.00019886383085385351, + "loss": 0.4949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49681541323661804, + "step": 1075, + "valid_targets_mean": 28777.3, + "valid_targets_min": 18981 + }, + { + "epoch": 1.152614727854856, + "grad_norm": 0.7444225763087141, + "learning_rate": 0.00019882856247916607, + "loss": 0.4939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.494374543428421, + "step": 1080, + "valid_targets_mean": 28757.2, + "valid_targets_min": 19339 + }, + { + "epoch": 1.1579509071504803, + "grad_norm": 0.6678571992056567, + "learning_rate": 0.00019879275827159486, + "loss": 0.4919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49021148681640625, + "step": 1085, + "valid_targets_mean": 28788.7, + "valid_targets_min": 21441 + }, + { + "epoch": 1.1632870864461047, + "grad_norm": 0.9789421150873826, + "learning_rate": 0.00019875641842526473, + "loss": 0.4913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49386441707611084, + "step": 1090, + "valid_targets_mean": 28768.3, + "valid_targets_min": 19307 + }, + { + "epoch": 1.1686232657417288, + "grad_norm": 0.7388801470880308, + "learning_rate": 0.00019871954313720454, + "loss": 0.4903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49126654863357544, + "step": 1095, + "valid_targets_mean": 28960.7, + "valid_targets_min": 20739 + }, + { + "epoch": 1.1739594450373532, + "grad_norm": 0.7313633738701135, + "learning_rate": 0.00019868213260734635, + "loss": 0.4921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4909404516220093, + "step": 1100, + "valid_targets_mean": 28717.3, + "valid_targets_min": 19151 + }, + { + "epoch": 1.1792956243329775, + "grad_norm": 0.6847916798801468, + "learning_rate": 0.00019864418703852406, + "loss": 0.4921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49161458015441895, + "step": 1105, + "valid_targets_mean": 28783.6, + "valid_targets_min": 18619 + }, + { + "epoch": 1.1846318036286019, + "grad_norm": 0.5992723218408188, + "learning_rate": 0.00019860570663647264, + "loss": 0.4931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49217694997787476, + "step": 1110, + "valid_targets_mean": 28852.7, + "valid_targets_min": 21815 + }, + { + "epoch": 1.1899679829242262, + "grad_norm": 0.7088694327208672, + "learning_rate": 0.00019856669160982674, + "loss": 0.4902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4884410500526428, + "step": 1115, + "valid_targets_mean": 28801.8, + "valid_targets_min": 20517 + }, + { + "epoch": 1.1953041622198506, + "grad_norm": 1.0128547253551068, + "learning_rate": 0.00019852714217011966, + "loss": 0.4901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4990355372428894, + "step": 1120, + "valid_targets_mean": 28929.9, + "valid_targets_min": 21103 + }, + { + "epoch": 1.200640341515475, + "grad_norm": 0.732187179688228, + "learning_rate": 0.0001984870585317823, + "loss": 0.4946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4941519498825073, + "step": 1125, + "valid_targets_mean": 28764.2, + "valid_targets_min": 21373 + }, + { + "epoch": 1.2059765208110993, + "grad_norm": 0.7655997856942812, + "learning_rate": 0.00019844644091214172, + "loss": 0.4918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49341729283332825, + "step": 1130, + "valid_targets_mean": 28565.8, + "valid_targets_min": 22833 + }, + { + "epoch": 1.2113127001067236, + "grad_norm": 0.6939402595770039, + "learning_rate": 0.00019840528953142032, + "loss": 0.4919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49373674392700195, + "step": 1135, + "valid_targets_mean": 29037.3, + "valid_targets_min": 21268 + }, + { + "epoch": 1.216648879402348, + "grad_norm": 0.8636563092762944, + "learning_rate": 0.0001983636046127344, + "loss": 0.4924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4902014136314392, + "step": 1140, + "valid_targets_mean": 28870.8, + "valid_targets_min": 19107 + }, + { + "epoch": 1.2219850586979724, + "grad_norm": 0.8594445582468951, + "learning_rate": 0.00019832138638209296, + "loss": 0.4895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4929233491420746, + "step": 1145, + "valid_targets_mean": 28872.9, + "valid_targets_min": 20602 + }, + { + "epoch": 1.2273212379935965, + "grad_norm": 0.7297644042655806, + "learning_rate": 0.00019827863506839666, + "loss": 0.4918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4909535050392151, + "step": 1150, + "valid_targets_mean": 28989.2, + "valid_targets_min": 21947 + }, + { + "epoch": 1.2326574172892208, + "grad_norm": 0.765568627312834, + "learning_rate": 0.00019823535090343636, + "loss": 0.4906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48845890164375305, + "step": 1155, + "valid_targets_mean": 28913.0, + "valid_targets_min": 20358 + }, + { + "epoch": 1.2379935965848452, + "grad_norm": 0.7025418654218685, + "learning_rate": 0.00019819153412189194, + "loss": 0.4877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48771578073501587, + "step": 1160, + "valid_targets_mean": 28797.3, + "valid_targets_min": 22345 + }, + { + "epoch": 1.2433297758804696, + "grad_norm": 0.5607803160935012, + "learning_rate": 0.00019814718496133108, + "loss": 0.4894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4898512363433838, + "step": 1165, + "valid_targets_mean": 28758.1, + "valid_targets_min": 17031 + }, + { + "epoch": 1.248665955176094, + "grad_norm": 0.9373306103102206, + "learning_rate": 0.00019810230366220803, + "loss": 0.4889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49357637763023376, + "step": 1170, + "valid_targets_mean": 28874.6, + "valid_targets_min": 22060 + }, + { + "epoch": 1.2540021344717183, + "grad_norm": 0.893301966708018, + "learning_rate": 0.00019805689046786202, + "loss": 0.4893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4874823987483978, + "step": 1175, + "valid_targets_mean": 28821.3, + "valid_targets_min": 16692 + }, + { + "epoch": 1.2593383137673426, + "grad_norm": 0.6848632438681702, + "learning_rate": 0.00019801094562451628, + "loss": 0.4904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49153780937194824, + "step": 1180, + "valid_targets_mean": 28801.0, + "valid_targets_min": 22226 + }, + { + "epoch": 1.264674493062967, + "grad_norm": 0.9103918804063192, + "learning_rate": 0.00019796446938127646, + "loss": 0.4893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4879952073097229, + "step": 1185, + "valid_targets_mean": 28990.6, + "valid_targets_min": 23411 + }, + { + "epoch": 1.2700106723585913, + "grad_norm": 0.7724803021110822, + "learning_rate": 0.00019791746199012948, + "loss": 0.49, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4941154718399048, + "step": 1190, + "valid_targets_mean": 28802.4, + "valid_targets_min": 22473 + }, + { + "epoch": 1.2753468516542155, + "grad_norm": 0.7173208410735338, + "learning_rate": 0.00019786992370594196, + "loss": 0.4868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4847111701965332, + "step": 1195, + "valid_targets_mean": 28934.8, + "valid_targets_min": 22545 + }, + { + "epoch": 1.2806830309498398, + "grad_norm": 0.9767771194720062, + "learning_rate": 0.00019782185478645902, + "loss": 0.4884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49301740527153015, + "step": 1200, + "valid_targets_mean": 28512.9, + "valid_targets_min": 21252 + }, + { + "epoch": 1.2860192102454642, + "grad_norm": 1.0781130531596468, + "learning_rate": 0.00019777325549230273, + "loss": 0.4878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49377936124801636, + "step": 1205, + "valid_targets_mean": 28668.4, + "valid_targets_min": 18606 + }, + { + "epoch": 1.2913553895410885, + "grad_norm": 0.8771564583626676, + "learning_rate": 0.00019772412608697079, + "loss": 0.4917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49229565262794495, + "step": 1210, + "valid_targets_mean": 28918.4, + "valid_targets_min": 21963 + }, + { + "epoch": 1.2966915688367129, + "grad_norm": 0.6869349410938043, + "learning_rate": 0.00019767446683683514, + "loss": 0.4907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4881063401699066, + "step": 1215, + "valid_targets_mean": 28948.3, + "valid_targets_min": 23020 + }, + { + "epoch": 1.3020277481323372, + "grad_norm": 0.6921187884299926, + "learning_rate": 0.00019762427801114037, + "loss": 0.4883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4854842722415924, + "step": 1220, + "valid_targets_mean": 28842.7, + "valid_targets_min": 19872 + }, + { + "epoch": 1.3073639274279616, + "grad_norm": 0.6841406169395504, + "learning_rate": 0.00019757355988200237, + "loss": 0.4874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.489562451839447, + "step": 1225, + "valid_targets_mean": 28766.4, + "valid_targets_min": 17167 + }, + { + "epoch": 1.312700106723586, + "grad_norm": 0.6106888444494654, + "learning_rate": 0.00019752231272440683, + "loss": 0.4863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4828183054924011, + "step": 1230, + "valid_targets_mean": 28774.1, + "valid_targets_min": 19426 + }, + { + "epoch": 1.3180362860192103, + "grad_norm": 0.5257233425597112, + "learning_rate": 0.00019747053681620785, + "loss": 0.4868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48759758472442627, + "step": 1235, + "valid_targets_mean": 28881.2, + "valid_targets_min": 23692 + }, + { + "epoch": 1.3233724653148347, + "grad_norm": 0.8480441282452694, + "learning_rate": 0.00019741823243812613, + "loss": 0.4885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48881375789642334, + "step": 1240, + "valid_targets_mean": 28882.3, + "valid_targets_min": 18555 + }, + { + "epoch": 1.328708644610459, + "grad_norm": 0.62182352123255, + "learning_rate": 0.0001973653998737478, + "loss": 0.4872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4828396439552307, + "step": 1245, + "valid_targets_mean": 28891.3, + "valid_targets_min": 21120 + }, + { + "epoch": 1.3340448239060834, + "grad_norm": 0.8004417732977367, + "learning_rate": 0.0001973120394095227, + "loss": 0.4843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4856739044189453, + "step": 1250, + "valid_targets_mean": 28789.2, + "valid_targets_min": 17800 + }, + { + "epoch": 1.3393810032017075, + "grad_norm": 0.7440408894557307, + "learning_rate": 0.00019725815133476288, + "loss": 0.4849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4856306314468384, + "step": 1255, + "valid_targets_mean": 28959.5, + "valid_targets_min": 22814 + }, + { + "epoch": 1.3447171824973319, + "grad_norm": 0.8407471815003433, + "learning_rate": 0.00019720373594164095, + "loss": 0.4854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48332279920578003, + "step": 1260, + "valid_targets_mean": 28746.5, + "valid_targets_min": 21515 + }, + { + "epoch": 1.3500533617929562, + "grad_norm": 0.5810747016865329, + "learning_rate": 0.00019714879352518858, + "loss": 0.4865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4885227680206299, + "step": 1265, + "valid_targets_mean": 28916.6, + "valid_targets_min": 22001 + }, + { + "epoch": 1.3553895410885806, + "grad_norm": 0.7600897840310242, + "learning_rate": 0.0001970933243832949, + "loss": 0.4859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4816964268684387, + "step": 1270, + "valid_targets_mean": 28778.2, + "valid_targets_min": 18530 + }, + { + "epoch": 1.360725720384205, + "grad_norm": 0.7625847143527156, + "learning_rate": 0.00019703732881670488, + "loss": 0.4847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4828633666038513, + "step": 1275, + "valid_targets_mean": 28908.7, + "valid_targets_min": 21741 + }, + { + "epoch": 1.3660618996798293, + "grad_norm": 0.7707967457925649, + "learning_rate": 0.00019698080712901765, + "loss": 0.4846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48737090826034546, + "step": 1280, + "valid_targets_mean": 28797.5, + "valid_targets_min": 20272 + }, + { + "epoch": 1.3713980789754536, + "grad_norm": 0.7118713515121713, + "learning_rate": 0.00019692375962668485, + "loss": 0.4835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4837408661842346, + "step": 1285, + "valid_targets_mean": 28756.1, + "valid_targets_min": 18040 + }, + { + "epoch": 1.376734258271078, + "grad_norm": 0.8007734251793294, + "learning_rate": 0.0001968661866190091, + "loss": 0.4848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.486257404088974, + "step": 1290, + "valid_targets_mean": 29051.0, + "valid_targets_min": 22393 + }, + { + "epoch": 1.3820704375667021, + "grad_norm": 0.707338396707381, + "learning_rate": 0.00019680808841814206, + "loss": 0.4823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4817184805870056, + "step": 1295, + "valid_targets_mean": 28809.5, + "valid_targets_min": 20321 + }, + { + "epoch": 1.3874066168623265, + "grad_norm": 0.6311339270049728, + "learning_rate": 0.00019674946533908315, + "loss": 0.482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4838714003562927, + "step": 1300, + "valid_targets_mean": 28615.7, + "valid_targets_min": 19155 + }, + { + "epoch": 1.3927427961579508, + "grad_norm": 0.7947663440225755, + "learning_rate": 0.0001966903176996774, + "loss": 0.4819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4826429486274719, + "step": 1305, + "valid_targets_mean": 28764.1, + "valid_targets_min": 19535 + }, + { + "epoch": 1.3980789754535752, + "grad_norm": 0.5399171640364682, + "learning_rate": 0.00019663064582061397, + "loss": 0.4828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48236802220344543, + "step": 1310, + "valid_targets_mean": 28867.7, + "valid_targets_min": 21662 + }, + { + "epoch": 1.4034151547491995, + "grad_norm": 0.7661910850810453, + "learning_rate": 0.00019657045002542442, + "loss": 0.4829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48537033796310425, + "step": 1315, + "valid_targets_mean": 28852.9, + "valid_targets_min": 18125 + }, + { + "epoch": 1.4087513340448239, + "grad_norm": 0.8066822149152647, + "learning_rate": 0.00019650973064048083, + "loss": 0.4808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48364531993865967, + "step": 1320, + "valid_targets_mean": 28965.1, + "valid_targets_min": 18255 + }, + { + "epoch": 1.4140875133404482, + "grad_norm": 0.6847873777978073, + "learning_rate": 0.00019644848799499413, + "loss": 0.4826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47989529371261597, + "step": 1325, + "valid_targets_mean": 28791.4, + "valid_targets_min": 17820 + }, + { + "epoch": 1.4194236926360726, + "grad_norm": 0.5872084117746706, + "learning_rate": 0.0001963867224210123, + "loss": 0.4823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48351019620895386, + "step": 1330, + "valid_targets_mean": 28846.1, + "valid_targets_min": 17670 + }, + { + "epoch": 1.424759871931697, + "grad_norm": 0.6656519836568731, + "learning_rate": 0.00019632443425341854, + "loss": 0.4831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.480230450630188, + "step": 1335, + "valid_targets_mean": 28764.0, + "valid_targets_min": 21322 + }, + { + "epoch": 1.4300960512273213, + "grad_norm": 0.588635393865431, + "learning_rate": 0.0001962616238299295, + "loss": 0.4806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4768039584159851, + "step": 1340, + "valid_targets_mean": 28781.9, + "valid_targets_min": 20991 + }, + { + "epoch": 1.4354322305229457, + "grad_norm": 0.6246083119855775, + "learning_rate": 0.00019619829149109336, + "loss": 0.4803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4811418652534485, + "step": 1345, + "valid_targets_mean": 28772.0, + "valid_targets_min": 17170 + }, + { + "epoch": 1.44076840981857, + "grad_norm": 0.7263321693572669, + "learning_rate": 0.0001961344375802881, + "loss": 0.4804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4802713990211487, + "step": 1350, + "valid_targets_mean": 28746.8, + "valid_targets_min": 19617 + }, + { + "epoch": 1.4461045891141944, + "grad_norm": 0.5996496688229794, + "learning_rate": 0.0001960700624437195, + "loss": 0.4798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47946688532829285, + "step": 1355, + "valid_targets_mean": 28772.0, + "valid_targets_min": 17857 + }, + { + "epoch": 1.4514407684098185, + "grad_norm": 0.5712796829856135, + "learning_rate": 0.00019600516643041943, + "loss": 0.4808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47957509756088257, + "step": 1360, + "valid_targets_mean": 28883.8, + "valid_targets_min": 19086 + }, + { + "epoch": 1.4567769477054429, + "grad_norm": 0.664555445537064, + "learning_rate": 0.0001959397498922439, + "loss": 0.4804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4794684052467346, + "step": 1365, + "valid_targets_mean": 28850.4, + "valid_targets_min": 18771 + }, + { + "epoch": 1.4621131270010672, + "grad_norm": 0.7318862747813851, + "learning_rate": 0.00019587381318387102, + "loss": 0.4767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4774578809738159, + "step": 1370, + "valid_targets_mean": 28921.8, + "valid_targets_min": 20968 + }, + { + "epoch": 1.4674493062966916, + "grad_norm": 0.7833483127423916, + "learning_rate": 0.0001958073566627992, + "loss": 0.4809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48020145297050476, + "step": 1375, + "valid_targets_mean": 28826.8, + "valid_targets_min": 20987 + }, + { + "epoch": 1.472785485592316, + "grad_norm": 0.8202494693452496, + "learning_rate": 0.00019574038068934525, + "loss": 0.4804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4827878773212433, + "step": 1380, + "valid_targets_mean": 28953.8, + "valid_targets_min": 22259 + }, + { + "epoch": 1.4781216648879403, + "grad_norm": 0.6624008136251276, + "learning_rate": 0.00019567288562664237, + "loss": 0.4818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4783748984336853, + "step": 1385, + "valid_targets_mean": 28843.0, + "valid_targets_min": 23178 + }, + { + "epoch": 1.4834578441835646, + "grad_norm": 0.7775036747710056, + "learning_rate": 0.00019560487184063806, + "loss": 0.4797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4827321767807007, + "step": 1390, + "valid_targets_mean": 29032.0, + "valid_targets_min": 22813 + }, + { + "epoch": 1.4887940234791888, + "grad_norm": 0.5649901315320555, + "learning_rate": 0.00019553633970009244, + "loss": 0.4802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48029249906539917, + "step": 1395, + "valid_targets_mean": 28964.1, + "valid_targets_min": 16470 + }, + { + "epoch": 1.4941302027748131, + "grad_norm": 0.5629363194657292, + "learning_rate": 0.00019546728957657588, + "loss": 0.4765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47325965762138367, + "step": 1400, + "valid_targets_mean": 29003.1, + "valid_targets_min": 23190 + }, + { + "epoch": 1.4994663820704375, + "grad_norm": 0.566425477316844, + "learning_rate": 0.00019539772184446735, + "loss": 0.4797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48466968536376953, + "step": 1405, + "valid_targets_mean": 28943.6, + "valid_targets_min": 18278 + }, + { + "epoch": 1.5048025613660618, + "grad_norm": 0.6745605040297462, + "learning_rate": 0.00019532763688095208, + "loss": 0.4772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4795975387096405, + "step": 1410, + "valid_targets_mean": 28871.0, + "valid_targets_min": 21906 + }, + { + "epoch": 1.5101387406616862, + "grad_norm": 0.7288603899967852, + "learning_rate": 0.0001952570350660197, + "loss": 0.4783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47614964842796326, + "step": 1415, + "valid_targets_mean": 28924.8, + "valid_targets_min": 16678 + }, + { + "epoch": 1.5154749199573105, + "grad_norm": 0.5587678214283813, + "learning_rate": 0.0001951859167824621, + "loss": 0.4796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47375237941741943, + "step": 1420, + "valid_targets_mean": 28987.3, + "valid_targets_min": 20241 + }, + { + "epoch": 1.520811099252935, + "grad_norm": 0.7688845619007809, + "learning_rate": 0.00019511428241587143, + "loss": 0.48, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4781056046485901, + "step": 1425, + "valid_targets_mean": 28988.3, + "valid_targets_min": 22679 + }, + { + "epoch": 1.5261472785485592, + "grad_norm": 0.5956710807064683, + "learning_rate": 0.00019504213235463792, + "loss": 0.4761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4784397482872009, + "step": 1430, + "valid_targets_mean": 28886.1, + "valid_targets_min": 22293 + }, + { + "epoch": 1.5314834578441836, + "grad_norm": 0.6641366454166832, + "learning_rate": 0.0001949694669899478, + "loss": 0.478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48540934920310974, + "step": 1435, + "valid_targets_mean": 28571.1, + "valid_targets_min": 21283 + }, + { + "epoch": 1.536819637139808, + "grad_norm": 0.7380686201977426, + "learning_rate": 0.00019489628671578126, + "loss": 0.478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47933629155158997, + "step": 1440, + "valid_targets_mean": 28792.6, + "valid_targets_min": 23082 + }, + { + "epoch": 1.5421558164354323, + "grad_norm": 0.762501599432968, + "learning_rate": 0.00019482259192891017, + "loss": 0.476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4763944149017334, + "step": 1445, + "valid_targets_mean": 28894.4, + "valid_targets_min": 19179 + }, + { + "epoch": 1.5474919957310567, + "grad_norm": 0.7870382961782788, + "learning_rate": 0.00019474838302889608, + "loss": 0.4753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4766830801963806, + "step": 1450, + "valid_targets_mean": 28894.8, + "valid_targets_min": 17014 + }, + { + "epoch": 1.552828175026681, + "grad_norm": 0.55480704737329, + "learning_rate": 0.00019467366041808797, + "loss": 0.4767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4800575375556946, + "step": 1455, + "valid_targets_mean": 28733.5, + "valid_targets_min": 22665 + }, + { + "epoch": 1.5581643543223054, + "grad_norm": 0.6590395456518542, + "learning_rate": 0.00019459842450161998, + "loss": 0.4783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4765167832374573, + "step": 1460, + "valid_targets_mean": 28797.2, + "valid_targets_min": 21481 + }, + { + "epoch": 1.5635005336179295, + "grad_norm": 0.644591706970926, + "learning_rate": 0.00019452267568740946, + "loss": 0.4758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47362422943115234, + "step": 1465, + "valid_targets_mean": 28885.0, + "valid_targets_min": 18996 + }, + { + "epoch": 1.5688367129135539, + "grad_norm": 0.641390008912236, + "learning_rate": 0.00019444641438615446, + "loss": 0.4767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4772486984729767, + "step": 1470, + "valid_targets_mean": 28653.0, + "valid_targets_min": 20800 + }, + { + "epoch": 1.5741728922091782, + "grad_norm": 0.5965323246358682, + "learning_rate": 0.00019436964101133178, + "loss": 0.4748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47263336181640625, + "step": 1475, + "valid_targets_mean": 28958.3, + "valid_targets_min": 21498 + }, + { + "epoch": 1.5795090715048026, + "grad_norm": 0.6746101269241358, + "learning_rate": 0.00019429235597919457, + "loss": 0.4743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46941155195236206, + "step": 1480, + "valid_targets_mean": 28767.7, + "valid_targets_min": 19146 + }, + { + "epoch": 1.584845250800427, + "grad_norm": 0.7539203708628796, + "learning_rate": 0.00019421455970877006, + "loss": 0.4748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4756488800048828, + "step": 1485, + "valid_targets_mean": 28706.9, + "valid_targets_min": 18866 + }, + { + "epoch": 1.590181430096051, + "grad_norm": 0.5581858891979666, + "learning_rate": 0.00019413625262185735, + "loss": 0.4772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47360649704933167, + "step": 1490, + "valid_targets_mean": 28828.8, + "valid_targets_min": 22508 + }, + { + "epoch": 1.5955176093916754, + "grad_norm": 1.1893212398925357, + "learning_rate": 0.00019405743514302516, + "loss": 0.4771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48227018117904663, + "step": 1495, + "valid_targets_mean": 28723.0, + "valid_targets_min": 16706 + }, + { + "epoch": 1.6008537886872998, + "grad_norm": 0.7594155355659317, + "learning_rate": 0.0001939781076996094, + "loss": 0.4758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47334688901901245, + "step": 1500, + "valid_targets_mean": 29004.8, + "valid_targets_min": 23444 + }, + { + "epoch": 1.6061899679829241, + "grad_norm": 0.6558435883064311, + "learning_rate": 0.00019389827072171096, + "loss": 0.4778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48009437322616577, + "step": 1505, + "valid_targets_mean": 28780.6, + "valid_targets_min": 20252 + }, + { + "epoch": 1.6115261472785485, + "grad_norm": 0.6462809800883502, + "learning_rate": 0.0001938179246421934, + "loss": 0.477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4776078760623932, + "step": 1510, + "valid_targets_mean": 29037.6, + "valid_targets_min": 19704 + }, + { + "epoch": 1.6168623265741728, + "grad_norm": 0.7619357108206934, + "learning_rate": 0.0001937370698966804, + "loss": 0.4749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4755183458328247, + "step": 1515, + "valid_targets_mean": 28866.1, + "valid_targets_min": 20479 + }, + { + "epoch": 1.6221985058697972, + "grad_norm": 0.6718256470924716, + "learning_rate": 0.00019365570692355373, + "loss": 0.476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47640013694763184, + "step": 1520, + "valid_targets_mean": 28887.7, + "valid_targets_min": 21952 + }, + { + "epoch": 1.6275346851654215, + "grad_norm": 0.5766744008183909, + "learning_rate": 0.00019357383616395055, + "loss": 0.4738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4738202691078186, + "step": 1525, + "valid_targets_mean": 28713.7, + "valid_targets_min": 20286 + }, + { + "epoch": 1.632870864461046, + "grad_norm": 0.5779758000586792, + "learning_rate": 0.00019349145806176125, + "loss": 0.4749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4797191023826599, + "step": 1530, + "valid_targets_mean": 28577.1, + "valid_targets_min": 20501 + }, + { + "epoch": 1.6382070437566703, + "grad_norm": 0.6742422191418099, + "learning_rate": 0.00019340857306362685, + "loss": 0.4736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47305184602737427, + "step": 1535, + "valid_targets_mean": 28852.1, + "valid_targets_min": 22994 + }, + { + "epoch": 1.6435432230522946, + "grad_norm": 0.588034899559399, + "learning_rate": 0.00019332518161893682, + "loss": 0.4747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47961777448654175, + "step": 1540, + "valid_targets_mean": 28754.7, + "valid_targets_min": 18503 + }, + { + "epoch": 1.648879402347919, + "grad_norm": 0.7222210515016969, + "learning_rate": 0.00019324128417982637, + "loss": 0.4734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4739217162132263, + "step": 1545, + "valid_targets_mean": 28845.9, + "valid_targets_min": 22369 + }, + { + "epoch": 1.6542155816435433, + "grad_norm": 0.7127517622189531, + "learning_rate": 0.0001931568812011742, + "loss": 0.4729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47446611523628235, + "step": 1550, + "valid_targets_mean": 28677.2, + "valid_targets_min": 18112 + }, + { + "epoch": 1.6595517609391677, + "grad_norm": 0.6484279198393836, + "learning_rate": 0.00019307197314059996, + "loss": 0.4741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4715951383113861, + "step": 1555, + "valid_targets_mean": 28785.7, + "valid_targets_min": 19884 + }, + { + "epoch": 1.664887940234792, + "grad_norm": 0.6384503419821512, + "learning_rate": 0.00019298656045846176, + "loss": 0.475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47511929273605347, + "step": 1560, + "valid_targets_mean": 28813.4, + "valid_targets_min": 22326 + }, + { + "epoch": 1.6702241195304164, + "grad_norm": 0.4962068369745076, + "learning_rate": 0.00019290064361785373, + "loss": 0.4735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4727150499820709, + "step": 1565, + "valid_targets_mean": 28866.5, + "valid_targets_min": 21069 + }, + { + "epoch": 1.6755602988260405, + "grad_norm": 0.6404261217666243, + "learning_rate": 0.00019281422308460336, + "loss": 0.4733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4757715165615082, + "step": 1570, + "valid_targets_mean": 29004.6, + "valid_targets_min": 23320 + }, + { + "epoch": 1.6808964781216649, + "grad_norm": 0.42810653136543797, + "learning_rate": 0.0001927272993272692, + "loss": 0.4726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4702845513820648, + "step": 1575, + "valid_targets_mean": 28811.0, + "valid_targets_min": 21827 + }, + { + "epoch": 1.6862326574172892, + "grad_norm": 0.5709891360685495, + "learning_rate": 0.00019263987281713818, + "loss": 0.4754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47373712062835693, + "step": 1580, + "valid_targets_mean": 28732.8, + "valid_targets_min": 20893 + }, + { + "epoch": 1.6915688367129136, + "grad_norm": 0.5834586825152454, + "learning_rate": 0.00019255194402822298, + "loss": 0.4722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47453397512435913, + "step": 1585, + "valid_targets_mean": 28772.9, + "valid_targets_min": 21085 + }, + { + "epoch": 1.696905016008538, + "grad_norm": 0.6470249725446496, + "learning_rate": 0.00019246351343725973, + "loss": 0.4711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.473699688911438, + "step": 1590, + "valid_targets_mean": 28835.1, + "valid_targets_min": 20580 + }, + { + "epoch": 1.702241195304162, + "grad_norm": 0.5973925947254088, + "learning_rate": 0.00019237458152370507, + "loss": 0.4706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4700833261013031, + "step": 1595, + "valid_targets_mean": 28884.6, + "valid_targets_min": 23013 + }, + { + "epoch": 1.7075773745997864, + "grad_norm": 0.5564214773287839, + "learning_rate": 0.00019228514876973386, + "loss": 0.4724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47462400794029236, + "step": 1600, + "valid_targets_mean": 28914.5, + "valid_targets_min": 22150 + }, + { + "epoch": 1.7129135538954108, + "grad_norm": 0.6833938785925946, + "learning_rate": 0.00019219521566023637, + "loss": 0.4703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4728323817253113, + "step": 1605, + "valid_targets_mean": 28960.5, + "valid_targets_min": 20794 + }, + { + "epoch": 1.7182497331910351, + "grad_norm": 0.5711738344386339, + "learning_rate": 0.00019210478268281576, + "loss": 0.4719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46786385774612427, + "step": 1610, + "valid_targets_mean": 28964.2, + "valid_targets_min": 21862 + }, + { + "epoch": 1.7235859124866595, + "grad_norm": 0.7651126866133595, + "learning_rate": 0.00019201385032778534, + "loss": 0.4748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47424450516700745, + "step": 1615, + "valid_targets_mean": 28854.0, + "valid_targets_min": 22067 + }, + { + "epoch": 1.7289220917822838, + "grad_norm": 0.7989209097175383, + "learning_rate": 0.00019192241908816602, + "loss": 0.4708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47150248289108276, + "step": 1620, + "valid_targets_mean": 28869.3, + "valid_targets_min": 20094 + }, + { + "epoch": 1.7342582710779082, + "grad_norm": 0.6681718819850506, + "learning_rate": 0.00019183048945968357, + "loss": 0.4705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4716719388961792, + "step": 1625, + "valid_targets_mean": 28579.5, + "valid_targets_min": 21541 + }, + { + "epoch": 1.7395944503735326, + "grad_norm": 0.5653742514176838, + "learning_rate": 0.00019173806194076597, + "loss": 0.4706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46959933638572693, + "step": 1630, + "valid_targets_mean": 28752.8, + "valid_targets_min": 19845 + }, + { + "epoch": 1.744930629669157, + "grad_norm": 0.5773186122160526, + "learning_rate": 0.0001916451370325406, + "loss": 0.4702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46714383363723755, + "step": 1635, + "valid_targets_mean": 29007.2, + "valid_targets_min": 20858 + }, + { + "epoch": 1.7502668089647813, + "grad_norm": 0.46286187641207616, + "learning_rate": 0.00019155171523883166, + "loss": 0.4698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46844184398651123, + "step": 1640, + "valid_targets_mean": 28928.1, + "valid_targets_min": 23272 + }, + { + "epoch": 1.7556029882604056, + "grad_norm": 0.6254757262158007, + "learning_rate": 0.00019145779706615745, + "loss": 0.4705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4707171618938446, + "step": 1645, + "valid_targets_mean": 28948.8, + "valid_targets_min": 21643 + }, + { + "epoch": 1.76093916755603, + "grad_norm": 0.6495021677221583, + "learning_rate": 0.00019136338302372746, + "loss": 0.4687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46310025453567505, + "step": 1650, + "valid_targets_mean": 28816.1, + "valid_targets_min": 22820 + }, + { + "epoch": 1.7662753468516543, + "grad_norm": 0.5538016735142929, + "learning_rate": 0.0001912684736234397, + "loss": 0.4684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4689071774482727, + "step": 1655, + "valid_targets_mean": 28719.9, + "valid_targets_min": 16202 + }, + { + "epoch": 1.7716115261472787, + "grad_norm": 0.6126692834218251, + "learning_rate": 0.00019117306937987803, + "loss": 0.4705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4690462350845337, + "step": 1660, + "valid_targets_mean": 28872.4, + "valid_targets_min": 19231 + }, + { + "epoch": 1.776947705442903, + "grad_norm": 0.6015035250124011, + "learning_rate": 0.00019107717081030918, + "loss": 0.4681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4632602632045746, + "step": 1665, + "valid_targets_mean": 28903.0, + "valid_targets_min": 20693 + }, + { + "epoch": 1.7822838847385272, + "grad_norm": 0.6310827401529582, + "learning_rate": 0.00019098077843468012, + "loss": 0.4685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47237634658813477, + "step": 1670, + "valid_targets_mean": 28865.4, + "valid_targets_min": 23113 + }, + { + "epoch": 1.7876200640341515, + "grad_norm": 0.6086750154228456, + "learning_rate": 0.00019088389277561508, + "loss": 0.469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46990033984184265, + "step": 1675, + "valid_targets_mean": 28757.7, + "valid_targets_min": 17338 + }, + { + "epoch": 1.7929562433297759, + "grad_norm": 0.6615357295397684, + "learning_rate": 0.00019078651435841285, + "loss": 0.4689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47078636288642883, + "step": 1680, + "valid_targets_mean": 28743.6, + "valid_targets_min": 21722 + }, + { + "epoch": 1.7982924226254002, + "grad_norm": 0.456475404398695, + "learning_rate": 0.0001906886437110438, + "loss": 0.4671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.465562641620636, + "step": 1685, + "valid_targets_mean": 28944.3, + "valid_targets_min": 22643 + }, + { + "epoch": 1.8036286019210246, + "grad_norm": 0.775533033267955, + "learning_rate": 0.0001905902813641472, + "loss": 0.4707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47184234857559204, + "step": 1690, + "valid_targets_mean": 28815.4, + "valid_targets_min": 22030 + }, + { + "epoch": 1.8089647812166487, + "grad_norm": 0.5641131554384016, + "learning_rate": 0.00019049142785102817, + "loss": 0.4692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47210821509361267, + "step": 1695, + "valid_targets_mean": 28887.3, + "valid_targets_min": 22019 + }, + { + "epoch": 1.814300960512273, + "grad_norm": 0.5982378230454685, + "learning_rate": 0.00019039208370765488, + "loss": 0.4676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4715844988822937, + "step": 1700, + "valid_targets_mean": 28786.1, + "valid_targets_min": 19559 + }, + { + "epoch": 1.8196371398078974, + "grad_norm": 0.7369184272659887, + "learning_rate": 0.00019029224947265561, + "loss": 0.4676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4698781967163086, + "step": 1705, + "valid_targets_mean": 28841.7, + "valid_targets_min": 18016 + }, + { + "epoch": 1.8249733191035218, + "grad_norm": 0.5760257831206574, + "learning_rate": 0.00019019192568731582, + "loss": 0.4673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4666833281517029, + "step": 1710, + "valid_targets_mean": 28801.2, + "valid_targets_min": 19500 + }, + { + "epoch": 1.8303094983991461, + "grad_norm": 0.6059236454872505, + "learning_rate": 0.0001900911128955753, + "loss": 0.4652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.466926634311676, + "step": 1715, + "valid_targets_mean": 28977.7, + "valid_targets_min": 22426 + }, + { + "epoch": 1.8356456776947705, + "grad_norm": 0.5727493374356408, + "learning_rate": 0.0001899898116440251, + "loss": 0.4654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.468447208404541, + "step": 1720, + "valid_targets_mean": 29006.2, + "valid_targets_min": 17008 + }, + { + "epoch": 1.8409818569903948, + "grad_norm": 0.5282364422069715, + "learning_rate": 0.00018988802248190458, + "loss": 0.4675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46732309460639954, + "step": 1725, + "valid_targets_mean": 28986.0, + "valid_targets_min": 22210 + }, + { + "epoch": 1.8463180362860192, + "grad_norm": 0.5911835364471943, + "learning_rate": 0.0001897857459610986, + "loss": 0.4668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4651775360107422, + "step": 1730, + "valid_targets_mean": 28794.5, + "valid_targets_min": 17770 + }, + { + "epoch": 1.8516542155816436, + "grad_norm": 0.6068041014668428, + "learning_rate": 0.0001896829826361343, + "loss": 0.4687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4736365079879761, + "step": 1735, + "valid_targets_mean": 28701.1, + "valid_targets_min": 20561 + }, + { + "epoch": 1.856990394877268, + "grad_norm": 0.6537224970166546, + "learning_rate": 0.0001895797330641782, + "loss": 0.4672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46415460109710693, + "step": 1740, + "valid_targets_mean": 28752.2, + "valid_targets_min": 21504 + }, + { + "epoch": 1.8623265741728923, + "grad_norm": 0.5992051872356972, + "learning_rate": 0.00018947599780503324, + "loss": 0.4661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46628502011299133, + "step": 1745, + "valid_targets_mean": 28908.7, + "valid_targets_min": 22966 + }, + { + "epoch": 1.8676627534685166, + "grad_norm": 0.5885838886628874, + "learning_rate": 0.0001893717774211356, + "loss": 0.4665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46606913208961487, + "step": 1750, + "valid_targets_mean": 29052.6, + "valid_targets_min": 21763 + }, + { + "epoch": 1.872998932764141, + "grad_norm": 0.7446688714076565, + "learning_rate": 0.0001892670724775518, + "loss": 0.4658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4657374620437622, + "step": 1755, + "valid_targets_mean": 28758.5, + "valid_targets_min": 21194 + }, + { + "epoch": 1.8783351120597653, + "grad_norm": 0.5036359705065248, + "learning_rate": 0.00018916188354197559, + "loss": 0.4654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.463339239358902, + "step": 1760, + "valid_targets_mean": 28803.6, + "valid_targets_min": 22932 + }, + { + "epoch": 1.8836712913553897, + "grad_norm": 0.5904829761655065, + "learning_rate": 0.00018905621118472468, + "loss": 0.4669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4687913954257965, + "step": 1765, + "valid_targets_mean": 28912.2, + "valid_targets_min": 21915 + }, + { + "epoch": 1.8890074706510138, + "grad_norm": 0.6758845024263211, + "learning_rate": 0.000188950055978738, + "loss": 0.4658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46433591842651367, + "step": 1770, + "valid_targets_mean": 28581.5, + "valid_targets_min": 17814 + }, + { + "epoch": 1.8943436499466382, + "grad_norm": 0.6655010490838106, + "learning_rate": 0.0001888434184995723, + "loss": 0.4678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47092902660369873, + "step": 1775, + "valid_targets_mean": 28813.9, + "valid_targets_min": 21894 + }, + { + "epoch": 1.8996798292422625, + "grad_norm": 0.6653183565853287, + "learning_rate": 0.00018873629932539921, + "loss": 0.4646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4640280604362488, + "step": 1780, + "valid_targets_mean": 28996.8, + "valid_targets_min": 22541 + }, + { + "epoch": 1.9050160085378869, + "grad_norm": 0.5578215869627684, + "learning_rate": 0.000188628699037002, + "loss": 0.4636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4635908007621765, + "step": 1785, + "valid_targets_mean": 28689.9, + "valid_targets_min": 19297 + }, + { + "epoch": 1.9103521878335112, + "grad_norm": 0.6785640133567981, + "learning_rate": 0.00018852061821777248, + "loss": 0.4647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4673336446285248, + "step": 1790, + "valid_targets_mean": 28844.9, + "valid_targets_min": 22033 + }, + { + "epoch": 1.9156883671291356, + "grad_norm": 0.49819218528460424, + "learning_rate": 0.00018841205745370775, + "loss": 0.4639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4637882113456726, + "step": 1795, + "valid_targets_mean": 28917.5, + "valid_targets_min": 20684 + }, + { + "epoch": 1.9210245464247597, + "grad_norm": 0.5735702051837462, + "learning_rate": 0.00018830301733340724, + "loss": 0.4657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4697621464729309, + "step": 1800, + "valid_targets_mean": 28733.1, + "valid_targets_min": 21774 + }, + { + "epoch": 1.926360725720384, + "grad_norm": 0.8086775803798638, + "learning_rate": 0.00018819349844806924, + "loss": 0.4659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4654462933540344, + "step": 1805, + "valid_targets_mean": 28812.4, + "valid_targets_min": 20848 + }, + { + "epoch": 1.9316969050160084, + "grad_norm": 0.763453791015228, + "learning_rate": 0.0001880835013914879, + "loss": 0.4619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4645582139492035, + "step": 1810, + "valid_targets_mean": 28802.2, + "valid_targets_min": 17513 + }, + { + "epoch": 1.9370330843116328, + "grad_norm": 0.6122778424943308, + "learning_rate": 0.00018797302676004988, + "loss": 0.4647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4665798246860504, + "step": 1815, + "valid_targets_mean": 28811.3, + "valid_targets_min": 22032 + }, + { + "epoch": 1.9423692636072571, + "grad_norm": 0.7904794382731611, + "learning_rate": 0.00018786207515273127, + "loss": 0.4668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46696364879608154, + "step": 1820, + "valid_targets_mean": 28671.8, + "valid_targets_min": 20916 + }, + { + "epoch": 1.9477054429028815, + "grad_norm": 0.6363143649888835, + "learning_rate": 0.00018775064717109415, + "loss": 0.4669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.465232789516449, + "step": 1825, + "valid_targets_mean": 28751.9, + "valid_targets_min": 21890 + }, + { + "epoch": 1.9530416221985059, + "grad_norm": 0.5872648061296175, + "learning_rate": 0.00018763874341928344, + "loss": 0.4645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4646969437599182, + "step": 1830, + "valid_targets_mean": 29092.1, + "valid_targets_min": 22338 + }, + { + "epoch": 1.9583778014941302, + "grad_norm": 0.5799013901167623, + "learning_rate": 0.00018752636450402373, + "loss": 0.4651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4642789959907532, + "step": 1835, + "valid_targets_mean": 28882.9, + "valid_targets_min": 19336 + }, + { + "epoch": 1.9637139807897546, + "grad_norm": 0.5552369882137852, + "learning_rate": 0.0001874135110346157, + "loss": 0.4637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4652862548828125, + "step": 1840, + "valid_targets_mean": 28639.8, + "valid_targets_min": 18392 + }, + { + "epoch": 1.969050160085379, + "grad_norm": 0.5877245255829862, + "learning_rate": 0.0001873001836229331, + "loss": 0.4651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4651973247528076, + "step": 1845, + "valid_targets_mean": 28827.9, + "valid_targets_min": 14728 + }, + { + "epoch": 1.9743863393810033, + "grad_norm": 0.6124274001415514, + "learning_rate": 0.00018718638288341933, + "loss": 0.4639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46294456720352173, + "step": 1850, + "valid_targets_mean": 28673.1, + "valid_targets_min": 19314 + }, + { + "epoch": 1.9797225186766276, + "grad_norm": 0.667045599764406, + "learning_rate": 0.00018707210943308397, + "loss": 0.4655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46378907561302185, + "step": 1855, + "valid_targets_mean": 28827.3, + "valid_targets_min": 23077 + }, + { + "epoch": 1.985058697972252, + "grad_norm": 0.570184918531401, + "learning_rate": 0.0001869573638914997, + "loss": 0.4623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45885172486305237, + "step": 1860, + "valid_targets_mean": 28967.2, + "valid_targets_min": 22321 + }, + { + "epoch": 1.9903948772678763, + "grad_norm": 0.7245936338973842, + "learning_rate": 0.00018684214688079875, + "loss": 0.4646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46353331208229065, + "step": 1865, + "valid_targets_mean": 28860.7, + "valid_targets_min": 22762 + }, + { + "epoch": 1.9957310565635007, + "grad_norm": 0.6639169072091015, + "learning_rate": 0.00018672645902566957, + "loss": 0.4623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46214547753334045, + "step": 1870, + "valid_targets_mean": 28939.7, + "valid_targets_min": 22612 + }, + { + "epoch": 2.001067235859125, + "grad_norm": 0.59933380977441, + "learning_rate": 0.0001866103009533535, + "loss": 0.4622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45977073907852173, + "step": 1875, + "valid_targets_mean": 28960.8, + "valid_targets_min": 23033 + }, + { + "epoch": 2.0064034151547494, + "grad_norm": 0.5187899325749421, + "learning_rate": 0.00018649367329364128, + "loss": 0.4608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4570675194263458, + "step": 1880, + "valid_targets_mean": 28719.8, + "valid_targets_min": 17665 + }, + { + "epoch": 2.0117395944503733, + "grad_norm": 0.5001063716442616, + "learning_rate": 0.00018637657667886965, + "loss": 0.4607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4625626802444458, + "step": 1885, + "valid_targets_mean": 28657.0, + "valid_targets_min": 17376 + }, + { + "epoch": 2.0170757737459977, + "grad_norm": 0.7323901122640898, + "learning_rate": 0.0001862590117439181, + "loss": 0.4599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46335369348526, + "step": 1890, + "valid_targets_mean": 28777.8, + "valid_targets_min": 21955 + }, + { + "epoch": 2.022411953041622, + "grad_norm": 0.5676814710375527, + "learning_rate": 0.00018614097912620506, + "loss": 0.4607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4586028456687927, + "step": 1895, + "valid_targets_mean": 28944.1, + "valid_targets_min": 22091 + }, + { + "epoch": 2.0277481323372464, + "grad_norm": 0.6827923425890666, + "learning_rate": 0.0001860224794656848, + "loss": 0.4587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45837873220443726, + "step": 1900, + "valid_targets_mean": 28785.9, + "valid_targets_min": 18288 + }, + { + "epoch": 2.0330843116328707, + "grad_norm": 0.600890619574247, + "learning_rate": 0.00018590351340484388, + "loss": 0.4595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4622270166873932, + "step": 1905, + "valid_targets_mean": 29004.6, + "valid_targets_min": 23067 + }, + { + "epoch": 2.038420490928495, + "grad_norm": 0.5654565589530285, + "learning_rate": 0.0001857840815886974, + "loss": 0.4583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45639216899871826, + "step": 1910, + "valid_targets_mean": 28861.2, + "valid_targets_min": 19565 + }, + { + "epoch": 2.0437566702241194, + "grad_norm": 0.6829281549445924, + "learning_rate": 0.0001856641846647859, + "loss": 0.4594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4606631398200989, + "step": 1915, + "valid_targets_mean": 28820.4, + "valid_targets_min": 18753 + }, + { + "epoch": 2.049092849519744, + "grad_norm": 0.5400921089118355, + "learning_rate": 0.00018554382328317163, + "loss": 0.4599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4611979126930237, + "step": 1920, + "valid_targets_mean": 28742.9, + "valid_targets_min": 22417 + }, + { + "epoch": 2.054429028815368, + "grad_norm": 0.6935582624005341, + "learning_rate": 0.00018542299809643493, + "loss": 0.4601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4632042348384857, + "step": 1925, + "valid_targets_mean": 28712.9, + "valid_targets_min": 18976 + }, + { + "epoch": 2.0597652081109925, + "grad_norm": 0.5813064771146723, + "learning_rate": 0.00018530170975967098, + "loss": 0.4604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.457609087228775, + "step": 1930, + "valid_targets_mean": 28734.9, + "valid_targets_min": 18178 + }, + { + "epoch": 2.065101387406617, + "grad_norm": 0.6538864203305739, + "learning_rate": 0.00018517995893048604, + "loss": 0.4604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46187877655029297, + "step": 1935, + "valid_targets_mean": 28943.5, + "valid_targets_min": 18084 + }, + { + "epoch": 2.070437566702241, + "grad_norm": 0.5842141291679069, + "learning_rate": 0.0001850577462689939, + "loss": 0.4588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45876073837280273, + "step": 1940, + "valid_targets_mean": 28907.3, + "valid_targets_min": 22624 + }, + { + "epoch": 2.0757737459978656, + "grad_norm": 0.6842902569703102, + "learning_rate": 0.00018493507243781224, + "loss": 0.4612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4589679539203644, + "step": 1945, + "valid_targets_mean": 28795.6, + "valid_targets_min": 22052 + }, + { + "epoch": 2.08110992529349, + "grad_norm": 0.635937259057549, + "learning_rate": 0.00018481193810205933, + "loss": 0.4586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.464557409286499, + "step": 1950, + "valid_targets_mean": 28863.1, + "valid_targets_min": 20012 + }, + { + "epoch": 2.0864461045891143, + "grad_norm": 0.5144324114257415, + "learning_rate": 0.00018468834392935016, + "loss": 0.4576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46036064624786377, + "step": 1955, + "valid_targets_mean": 28950.8, + "valid_targets_min": 22641 + }, + { + "epoch": 2.0917822838847386, + "grad_norm": 0.5987241157449255, + "learning_rate": 0.00018456429058979278, + "loss": 0.4601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4588262736797333, + "step": 1960, + "valid_targets_mean": 28777.1, + "valid_targets_min": 19931 + }, + { + "epoch": 2.097118463180363, + "grad_norm": 0.6866158518403197, + "learning_rate": 0.0001844397787559848, + "loss": 0.4574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.457817018032074, + "step": 1965, + "valid_targets_mean": 28779.0, + "valid_targets_min": 19641 + }, + { + "epoch": 2.1024546424759873, + "grad_norm": 0.5291979866110622, + "learning_rate": 0.00018431480910300984, + "loss": 0.456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45642563700675964, + "step": 1970, + "valid_targets_mean": 28720.2, + "valid_targets_min": 21564 + }, + { + "epoch": 2.1077908217716117, + "grad_norm": 0.6023975429951549, + "learning_rate": 0.00018418938230843358, + "loss": 0.4578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45973849296569824, + "step": 1975, + "valid_targets_mean": 28830.4, + "valid_targets_min": 21782 + }, + { + "epoch": 2.113127001067236, + "grad_norm": 0.5115658188796195, + "learning_rate": 0.00018406349905230037, + "loss": 0.4581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4587351083755493, + "step": 1980, + "valid_targets_mean": 28822.9, + "valid_targets_min": 18936 + }, + { + "epoch": 2.11846318036286, + "grad_norm": 0.7055463833686328, + "learning_rate": 0.0001839371600171294, + "loss": 0.4587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4615404009819031, + "step": 1985, + "valid_targets_mean": 28784.8, + "valid_targets_min": 19860 + }, + { + "epoch": 2.1237993596584843, + "grad_norm": 0.6445978890910325, + "learning_rate": 0.00018381036588791094, + "loss": 0.4591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46140262484550476, + "step": 1990, + "valid_targets_mean": 28812.8, + "valid_targets_min": 20617 + }, + { + "epoch": 2.1291355389541087, + "grad_norm": 0.5642093421556056, + "learning_rate": 0.0001836831173521028, + "loss": 0.4587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45539408922195435, + "step": 1995, + "valid_targets_mean": 28769.3, + "valid_targets_min": 23476 + }, + { + "epoch": 2.134471718249733, + "grad_norm": 0.5633066242804747, + "learning_rate": 0.00018355541509962645, + "loss": 0.4567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4613940715789795, + "step": 2000, + "valid_targets_mean": 28864.9, + "valid_targets_min": 20137 + }, + { + "epoch": 2.1398078975453574, + "grad_norm": 0.7423375483745515, + "learning_rate": 0.00018342725982286342, + "loss": 0.4576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45607930421829224, + "step": 2005, + "valid_targets_mean": 28840.3, + "valid_targets_min": 19525 + }, + { + "epoch": 2.1451440768409817, + "grad_norm": 0.6407532064844419, + "learning_rate": 0.00018329865221665138, + "loss": 0.4579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.46243995428085327, + "step": 2010, + "valid_targets_mean": 28735.4, + "valid_targets_min": 21423 + }, + { + "epoch": 2.150480256136606, + "grad_norm": 0.5071631414893789, + "learning_rate": 0.0001831695929782805, + "loss": 0.4571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4553641676902771, + "step": 2015, + "valid_targets_mean": 28805.2, + "valid_targets_min": 21289 + }, + { + "epoch": 2.1558164354322304, + "grad_norm": 0.5351424501446503, + "learning_rate": 0.00018304008280748963, + "loss": 0.4577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45526012778282166, + "step": 2020, + "valid_targets_mean": 28898.0, + "valid_targets_min": 22334 + }, + { + "epoch": 2.161152614727855, + "grad_norm": 0.5832349897737564, + "learning_rate": 0.00018291012240646246, + "loss": 0.4588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4605897068977356, + "step": 2025, + "valid_targets_mean": 28928.8, + "valid_targets_min": 22394 + }, + { + "epoch": 2.166488794023479, + "grad_norm": 0.6341432814176831, + "learning_rate": 0.00018277971247982383, + "loss": 0.4576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45824578404426575, + "step": 2030, + "valid_targets_mean": 28848.5, + "valid_targets_min": 23009 + }, + { + "epoch": 2.1718249733191035, + "grad_norm": 0.5618988276078262, + "learning_rate": 0.0001826488537346358, + "loss": 0.4549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4593442380428314, + "step": 2035, + "valid_targets_mean": 28906.0, + "valid_targets_min": 22415 + }, + { + "epoch": 2.177161152614728, + "grad_norm": 0.5852316676215777, + "learning_rate": 0.00018251754688039385, + "loss": 0.4565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45731204748153687, + "step": 2040, + "valid_targets_mean": 28754.1, + "valid_targets_min": 21083 + }, + { + "epoch": 2.1824973319103522, + "grad_norm": 0.5470934333926433, + "learning_rate": 0.00018238579262902301, + "loss": 0.4566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45632243156433105, + "step": 2045, + "valid_targets_mean": 28928.7, + "valid_targets_min": 19931 + }, + { + "epoch": 2.1878335112059766, + "grad_norm": 0.5415715818158156, + "learning_rate": 0.00018225359169487412, + "loss": 0.4568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4580104947090149, + "step": 2050, + "valid_targets_mean": 28765.1, + "valid_targets_min": 20572 + }, + { + "epoch": 2.193169690501601, + "grad_norm": 0.598303068188444, + "learning_rate": 0.00018212094479471975, + "loss": 0.457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45613572001457214, + "step": 2055, + "valid_targets_mean": 28840.7, + "valid_targets_min": 21709 + }, + { + "epoch": 2.1985058697972253, + "grad_norm": 0.6913200758318999, + "learning_rate": 0.0001819878526477505, + "loss": 0.4567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45595425367355347, + "step": 2060, + "valid_targets_mean": 28777.8, + "valid_targets_min": 22706 + }, + { + "epoch": 2.2038420490928496, + "grad_norm": 0.40806618165354064, + "learning_rate": 0.000181854315975571, + "loss": 0.4561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4540978968143463, + "step": 2065, + "valid_targets_mean": 28891.5, + "valid_targets_min": 21723 + }, + { + "epoch": 2.209178228388474, + "grad_norm": 0.6338950455728084, + "learning_rate": 0.000181720335502196, + "loss": 0.4552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45749494433403015, + "step": 2070, + "valid_targets_mean": 28764.5, + "valid_targets_min": 19535 + }, + { + "epoch": 2.2145144076840984, + "grad_norm": 0.5615674622004101, + "learning_rate": 0.0001815859119540466, + "loss": 0.4561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4547135829925537, + "step": 2075, + "valid_targets_mean": 29034.3, + "valid_targets_min": 20190 + }, + { + "epoch": 2.2198505869797227, + "grad_norm": 0.5942171627491899, + "learning_rate": 0.00018145104605994599, + "loss": 0.459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45587158203125, + "step": 2080, + "valid_targets_mean": 28982.1, + "valid_targets_min": 23244 + }, + { + "epoch": 2.225186766275347, + "grad_norm": 0.5544783734850327, + "learning_rate": 0.00018131573855111578, + "loss": 0.4563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45724672079086304, + "step": 2085, + "valid_targets_mean": 28879.3, + "valid_targets_min": 19495 + }, + { + "epoch": 2.2305229455709714, + "grad_norm": 0.6646252391429057, + "learning_rate": 0.00018117999016117203, + "loss": 0.4556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4570236802101135, + "step": 2090, + "valid_targets_mean": 28734.1, + "valid_targets_min": 19166 + }, + { + "epoch": 2.2358591248665953, + "grad_norm": 0.6369862489589629, + "learning_rate": 0.00018104380162612101, + "loss": 0.4565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4539600908756256, + "step": 2095, + "valid_targets_mean": 28860.5, + "valid_targets_min": 22225 + }, + { + "epoch": 2.2411953041622197, + "grad_norm": 0.6131804501294443, + "learning_rate": 0.0001809071736843556, + "loss": 0.4555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4549659490585327, + "step": 2100, + "valid_targets_mean": 28904.1, + "valid_targets_min": 22427 + }, + { + "epoch": 2.246531483457844, + "grad_norm": 0.5523653824690307, + "learning_rate": 0.00018077010707665084, + "loss": 0.4549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.455466628074646, + "step": 2105, + "valid_targets_mean": 28753.7, + "valid_targets_min": 23155 + }, + { + "epoch": 2.2518676627534684, + "grad_norm": 0.5728452825342911, + "learning_rate": 0.00018063260254616033, + "loss": 0.4563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4555572271347046, + "step": 2110, + "valid_targets_mean": 28900.0, + "valid_targets_min": 21903 + }, + { + "epoch": 2.2572038420490927, + "grad_norm": 0.7136693673051113, + "learning_rate": 0.00018049466083841194, + "loss": 0.4558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45720332860946655, + "step": 2115, + "valid_targets_mean": 28775.8, + "valid_targets_min": 22119 + }, + { + "epoch": 2.262540021344717, + "grad_norm": 0.4503189370802511, + "learning_rate": 0.0001803562827013039, + "loss": 0.4546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45119166374206543, + "step": 2120, + "valid_targets_mean": 28674.5, + "valid_targets_min": 20689 + }, + { + "epoch": 2.2678762006403415, + "grad_norm": 0.5722858990868676, + "learning_rate": 0.00018021746888510057, + "loss": 0.4576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4569234848022461, + "step": 2125, + "valid_targets_mean": 28806.0, + "valid_targets_min": 20865 + }, + { + "epoch": 2.273212379935966, + "grad_norm": 0.5096040586782076, + "learning_rate": 0.00018007822014242867, + "loss": 0.4553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4586583971977234, + "step": 2130, + "valid_targets_mean": 28771.0, + "valid_targets_min": 22431 + }, + { + "epoch": 2.27854855923159, + "grad_norm": 0.5617317894350805, + "learning_rate": 0.00017993853722827284, + "loss": 0.4544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4546637535095215, + "step": 2135, + "valid_targets_mean": 28571.8, + "valid_targets_min": 20067 + }, + { + "epoch": 2.2838847385272145, + "grad_norm": 0.5270533340074951, + "learning_rate": 0.00017979842089997186, + "loss": 0.4561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.452915757894516, + "step": 2140, + "valid_targets_mean": 28877.1, + "valid_targets_min": 20734 + }, + { + "epoch": 2.289220917822839, + "grad_norm": 0.4882137321818793, + "learning_rate": 0.00017965787191721435, + "loss": 0.4558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45327115058898926, + "step": 2145, + "valid_targets_mean": 28830.8, + "valid_targets_min": 18590 + }, + { + "epoch": 2.2945570971184632, + "grad_norm": 0.6423744476600158, + "learning_rate": 0.0001795168910420348, + "loss": 0.4527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4503288269042969, + "step": 2150, + "valid_targets_mean": 28775.4, + "valid_targets_min": 20912 + }, + { + "epoch": 2.2998932764140876, + "grad_norm": 0.6490306301093351, + "learning_rate": 0.00017937547903880917, + "loss": 0.4544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.455191969871521, + "step": 2155, + "valid_targets_mean": 28614.9, + "valid_targets_min": 22354 + }, + { + "epoch": 2.305229455709712, + "grad_norm": 0.5582032512637555, + "learning_rate": 0.00017923363667425116, + "loss": 0.4529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45650357007980347, + "step": 2160, + "valid_targets_mean": 28915.7, + "valid_targets_min": 18988 + }, + { + "epoch": 2.3105656350053363, + "grad_norm": 0.6009137653662315, + "learning_rate": 0.00017909136471740765, + "loss": 0.4552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4560841917991638, + "step": 2165, + "valid_targets_mean": 28753.2, + "valid_targets_min": 21835 + }, + { + "epoch": 2.3159018143009606, + "grad_norm": 0.6041106978318558, + "learning_rate": 0.00017894866393965476, + "loss": 0.4539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45484206080436707, + "step": 2170, + "valid_targets_mean": 28878.6, + "valid_targets_min": 22043 + }, + { + "epoch": 2.321237993596585, + "grad_norm": 0.4272191457749068, + "learning_rate": 0.0001788055351146936, + "loss": 0.4543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4543980062007904, + "step": 2175, + "valid_targets_mean": 28866.5, + "valid_targets_min": 21992 + }, + { + "epoch": 2.3265741728922094, + "grad_norm": 0.6502688723355644, + "learning_rate": 0.00017866197901854615, + "loss": 0.455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45570501685142517, + "step": 2180, + "valid_targets_mean": 28962.6, + "valid_targets_min": 19883 + }, + { + "epoch": 2.3319103521878333, + "grad_norm": 0.4024183500508293, + "learning_rate": 0.00017851799642955088, + "loss": 0.4568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4559852182865143, + "step": 2185, + "valid_targets_mean": 28797.3, + "valid_targets_min": 22052 + }, + { + "epoch": 2.3372465314834576, + "grad_norm": 0.4693015865238889, + "learning_rate": 0.00017837358812835866, + "loss": 0.457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4545251131057739, + "step": 2190, + "valid_targets_mean": 28810.5, + "valid_targets_min": 19893 + }, + { + "epoch": 2.342582710779082, + "grad_norm": 0.5223725512693111, + "learning_rate": 0.0001782287548979286, + "loss": 0.4542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4560278654098511, + "step": 2195, + "valid_targets_mean": 28810.0, + "valid_targets_min": 21589 + }, + { + "epoch": 2.3479188900747063, + "grad_norm": 0.5178388414730264, + "learning_rate": 0.00017808349752352357, + "loss": 0.4551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4541753828525543, + "step": 2200, + "valid_targets_mean": 28903.5, + "valid_targets_min": 17794 + }, + { + "epoch": 2.3532550693703307, + "grad_norm": 0.5484107517474188, + "learning_rate": 0.0001779378167927062, + "loss": 0.4553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4518164098262787, + "step": 2205, + "valid_targets_mean": 29016.5, + "valid_targets_min": 22577 + }, + { + "epoch": 2.358591248665955, + "grad_norm": 0.5707202376642307, + "learning_rate": 0.00017779171349533446, + "loss": 0.4542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.453921377658844, + "step": 2210, + "valid_targets_mean": 28946.0, + "valid_targets_min": 22579 + }, + { + "epoch": 2.3639274279615794, + "grad_norm": 0.6507332395005175, + "learning_rate": 0.0001776451884235573, + "loss": 0.4542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45342642068862915, + "step": 2215, + "valid_targets_mean": 28768.5, + "valid_targets_min": 19660 + }, + { + "epoch": 2.3692636072572038, + "grad_norm": 0.5253254081733058, + "learning_rate": 0.00017749824237181066, + "loss": 0.4528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.454387366771698, + "step": 2220, + "valid_targets_mean": 28863.5, + "valid_targets_min": 21233 + }, + { + "epoch": 2.374599786552828, + "grad_norm": 0.5749454692936622, + "learning_rate": 0.00017735087613681284, + "loss": 0.4535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4516810178756714, + "step": 2225, + "valid_targets_mean": 28899.3, + "valid_targets_min": 21644 + }, + { + "epoch": 2.3799359658484525, + "grad_norm": 0.5333442269729424, + "learning_rate": 0.00017720309051756026, + "loss": 0.4537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45168444514274597, + "step": 2230, + "valid_targets_mean": 28924.2, + "valid_targets_min": 22883 + }, + { + "epoch": 2.385272145144077, + "grad_norm": 0.5804156748991731, + "learning_rate": 0.00017705488631532333, + "loss": 0.4532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4527837634086609, + "step": 2235, + "valid_targets_mean": 28909.1, + "valid_targets_min": 21999 + }, + { + "epoch": 2.390608324439701, + "grad_norm": 0.5369453418155822, + "learning_rate": 0.00017690626433364184, + "loss": 0.4504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4510282874107361, + "step": 2240, + "valid_targets_mean": 28867.9, + "valid_targets_min": 19981 + }, + { + "epoch": 2.3959445037353255, + "grad_norm": 0.5695614019290133, + "learning_rate": 0.00017675722537832073, + "loss": 0.4545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4558982253074646, + "step": 2245, + "valid_targets_mean": 28881.5, + "valid_targets_min": 21356 + }, + { + "epoch": 2.40128068303095, + "grad_norm": 0.44039244965159946, + "learning_rate": 0.00017660777025742572, + "loss": 0.4526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45342639088630676, + "step": 2250, + "valid_targets_mean": 28929.8, + "valid_targets_min": 20524 + }, + { + "epoch": 2.4066168623265742, + "grad_norm": 0.4152414652406428, + "learning_rate": 0.00017645789978127893, + "loss": 0.4519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45057374238967896, + "step": 2255, + "valid_targets_mean": 28970.9, + "valid_targets_min": 20859 + }, + { + "epoch": 2.4119530416221986, + "grad_norm": 0.6508404516893584, + "learning_rate": 0.00017630761476245446, + "loss": 0.4541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4561172127723694, + "step": 2260, + "valid_targets_mean": 28894.7, + "valid_targets_min": 23451 + }, + { + "epoch": 2.417289220917823, + "grad_norm": 0.5170269358755798, + "learning_rate": 0.00017615691601577393, + "loss": 0.4538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45423901081085205, + "step": 2265, + "valid_targets_mean": 28926.5, + "valid_targets_min": 21125 + }, + { + "epoch": 2.4226254002134473, + "grad_norm": 0.5539052167798649, + "learning_rate": 0.00017600580435830226, + "loss": 0.4539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4492901563644409, + "step": 2270, + "valid_targets_mean": 29171.7, + "valid_targets_min": 18853 + }, + { + "epoch": 2.4279615795090717, + "grad_norm": 0.5027064755030347, + "learning_rate": 0.00017585428060934295, + "loss": 0.4513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44866707921028137, + "step": 2275, + "valid_targets_mean": 28943.6, + "valid_targets_min": 20915 + }, + { + "epoch": 2.433297758804696, + "grad_norm": 0.49172724471201934, + "learning_rate": 0.00017570234559043392, + "loss": 0.4533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44804149866104126, + "step": 2280, + "valid_targets_mean": 28720.4, + "valid_targets_min": 17376 + }, + { + "epoch": 2.4386339381003204, + "grad_norm": 0.5303285183291916, + "learning_rate": 0.00017555000012534292, + "loss": 0.4509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4553695321083069, + "step": 2285, + "valid_targets_mean": 28826.4, + "valid_targets_min": 19118 + }, + { + "epoch": 2.4439701173959447, + "grad_norm": 0.4350864203473741, + "learning_rate": 0.000175397245040063, + "loss": 0.4542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4559476375579834, + "step": 2290, + "valid_targets_mean": 28908.8, + "valid_targets_min": 22187 + }, + { + "epoch": 2.449306296691569, + "grad_norm": 0.6127096293789136, + "learning_rate": 0.00017524408116280813, + "loss": 0.4533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45513585209846497, + "step": 2295, + "valid_targets_mean": 28868.0, + "valid_targets_min": 15207 + }, + { + "epoch": 2.454642475987193, + "grad_norm": 0.5051617324594502, + "learning_rate": 0.00017509050932400876, + "loss": 0.4529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4498024880886078, + "step": 2300, + "valid_targets_mean": 28820.4, + "valid_targets_min": 20899 + }, + { + "epoch": 2.4599786552828173, + "grad_norm": 0.5619716934988601, + "learning_rate": 0.00017493653035630715, + "loss": 0.4529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.451460599899292, + "step": 2305, + "valid_targets_mean": 28765.6, + "valid_targets_min": 20019 + }, + { + "epoch": 2.4653148345784417, + "grad_norm": 0.49243931962655774, + "learning_rate": 0.000174782145094553, + "loss": 0.4506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4520164132118225, + "step": 2310, + "valid_targets_mean": 28843.5, + "valid_targets_min": 17007 + }, + { + "epoch": 2.470651013874066, + "grad_norm": 0.582177272434108, + "learning_rate": 0.00017462735437579884, + "loss": 0.4544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4531388580799103, + "step": 2315, + "valid_targets_mean": 28881.2, + "valid_targets_min": 22525 + }, + { + "epoch": 2.4759871931696904, + "grad_norm": 0.6116841941244345, + "learning_rate": 0.00017447215903929552, + "loss": 0.452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4520295560359955, + "step": 2320, + "valid_targets_mean": 28662.7, + "valid_targets_min": 21514 + }, + { + "epoch": 2.4813233724653148, + "grad_norm": 0.5184075672422216, + "learning_rate": 0.00017431655992648776, + "loss": 0.452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45252466201782227, + "step": 2325, + "valid_targets_mean": 28927.5, + "valid_targets_min": 23202 + }, + { + "epoch": 2.486659551760939, + "grad_norm": 0.6047406574094899, + "learning_rate": 0.00017416055788100935, + "loss": 0.4517, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45016252994537354, + "step": 2330, + "valid_targets_mean": 28785.7, + "valid_targets_min": 21020 + }, + { + "epoch": 2.4919957310565635, + "grad_norm": 0.5824136089989772, + "learning_rate": 0.0001740041537486788, + "loss": 0.4517, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4471256136894226, + "step": 2335, + "valid_targets_mean": 29024.5, + "valid_targets_min": 23134 + }, + { + "epoch": 2.497331910352188, + "grad_norm": 0.5065556408907309, + "learning_rate": 0.00017384734837749472, + "loss": 0.4505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4542456865310669, + "step": 2340, + "valid_targets_mean": 28765.2, + "valid_targets_min": 21832 + }, + { + "epoch": 2.502668089647812, + "grad_norm": 0.552973006999218, + "learning_rate": 0.000173690142617631, + "loss": 0.4509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4535852074623108, + "step": 2345, + "valid_targets_mean": 28844.7, + "valid_targets_min": 19678 + }, + { + "epoch": 2.5080042689434365, + "grad_norm": 0.4743900747981737, + "learning_rate": 0.00017353253732143255, + "loss": 0.4518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4519196152687073, + "step": 2350, + "valid_targets_mean": 28954.9, + "valid_targets_min": 22529 + }, + { + "epoch": 2.513340448239061, + "grad_norm": 0.562657221120804, + "learning_rate": 0.00017337453334341044, + "loss": 0.4512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44621649384498596, + "step": 2355, + "valid_targets_mean": 28777.2, + "valid_targets_min": 22850 + }, + { + "epoch": 2.5186766275346852, + "grad_norm": 0.6273755858177126, + "learning_rate": 0.00017321613154023727, + "loss": 0.4513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4545447826385498, + "step": 2360, + "valid_targets_mean": 28767.7, + "valid_targets_min": 22860 + }, + { + "epoch": 2.5240128068303096, + "grad_norm": 0.5944845566511393, + "learning_rate": 0.00017305733277074272, + "loss": 0.4495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4476246237754822, + "step": 2365, + "valid_targets_mean": 28874.1, + "valid_targets_min": 16924 + }, + { + "epoch": 2.529348986125934, + "grad_norm": 0.44790911582263976, + "learning_rate": 0.00017289813789590864, + "loss": 0.4522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4487800896167755, + "step": 2370, + "valid_targets_mean": 28719.8, + "valid_targets_min": 20685 + }, + { + "epoch": 2.5346851654215583, + "grad_norm": 0.743814817864195, + "learning_rate": 0.0001727385477788645, + "loss": 0.4505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4476528763771057, + "step": 2375, + "valid_targets_mean": 29018.1, + "valid_targets_min": 22291 + }, + { + "epoch": 2.5400213447171827, + "grad_norm": 0.5011954170333152, + "learning_rate": 0.00017257856328488285, + "loss": 0.4531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45606738328933716, + "step": 2380, + "valid_targets_mean": 28651.0, + "valid_targets_min": 21448 + }, + { + "epoch": 2.5453575240128066, + "grad_norm": 0.48315014269896034, + "learning_rate": 0.0001724181852813743, + "loss": 0.4503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44993042945861816, + "step": 2385, + "valid_targets_mean": 28859.8, + "valid_targets_min": 21386 + }, + { + "epoch": 2.550693703308431, + "grad_norm": 0.49675712419607215, + "learning_rate": 0.00017225741463788315, + "loss": 0.4505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45400500297546387, + "step": 2390, + "valid_targets_mean": 28968.6, + "valid_targets_min": 17603 + }, + { + "epoch": 2.5560298826040553, + "grad_norm": 0.5173302426897215, + "learning_rate": 0.0001720962522260825, + "loss": 0.4521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45111632347106934, + "step": 2395, + "valid_targets_mean": 28820.2, + "valid_targets_min": 19026 + }, + { + "epoch": 2.5613660618996796, + "grad_norm": 0.6860815784203482, + "learning_rate": 0.00017193469891976947, + "loss": 0.4494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4513351619243622, + "step": 2400, + "valid_targets_mean": 28871.1, + "valid_targets_min": 22025 + }, + { + "epoch": 2.566702241195304, + "grad_norm": 0.7030866923646217, + "learning_rate": 0.00017177275559486068, + "loss": 0.4521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4503113627433777, + "step": 2405, + "valid_targets_mean": 28827.6, + "valid_targets_min": 19319 + }, + { + "epoch": 2.5720384204909283, + "grad_norm": 0.43516321280201453, + "learning_rate": 0.00017161042312938723, + "loss": 0.4509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4512559771537781, + "step": 2410, + "valid_targets_mean": 28753.2, + "valid_targets_min": 22762 + }, + { + "epoch": 2.5773745997865527, + "grad_norm": 0.5133859107024392, + "learning_rate": 0.00017144770240349022, + "loss": 0.45, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4520412087440491, + "step": 2415, + "valid_targets_mean": 28935.5, + "valid_targets_min": 21925 + }, + { + "epoch": 2.582710779082177, + "grad_norm": 0.5064856540613286, + "learning_rate": 0.00017128459429941576, + "loss": 0.4485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44660258293151855, + "step": 2420, + "valid_targets_mean": 28878.1, + "valid_targets_min": 18457 + }, + { + "epoch": 2.5880469583778014, + "grad_norm": 0.5846486099782726, + "learning_rate": 0.00017112109970151025, + "loss": 0.4516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45156729221343994, + "step": 2425, + "valid_targets_mean": 28922.7, + "valid_targets_min": 19523 + }, + { + "epoch": 2.5933831376734258, + "grad_norm": 0.5187420506239093, + "learning_rate": 0.00017095721949621565, + "loss": 0.4486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4488029181957245, + "step": 2430, + "valid_targets_mean": 28784.3, + "valid_targets_min": 18489 + }, + { + "epoch": 2.59871931696905, + "grad_norm": 0.5900113475966927, + "learning_rate": 0.00017079295457206456, + "loss": 0.4504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4478583335876465, + "step": 2435, + "valid_targets_mean": 28929.3, + "valid_targets_min": 20301 + }, + { + "epoch": 2.6040554962646745, + "grad_norm": 0.5132399927767499, + "learning_rate": 0.0001706283058196755, + "loss": 0.448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44799095392227173, + "step": 2440, + "valid_targets_mean": 28912.8, + "valid_targets_min": 20144 + }, + { + "epoch": 2.609391675560299, + "grad_norm": 0.4623662629357846, + "learning_rate": 0.0001704632741317481, + "loss": 0.4482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4481392502784729, + "step": 2445, + "valid_targets_mean": 28754.9, + "valid_targets_min": 22290 + }, + { + "epoch": 2.614727854855923, + "grad_norm": 0.5131007295366438, + "learning_rate": 0.00017029786040305807, + "loss": 0.4498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44938600063323975, + "step": 2450, + "valid_targets_mean": 28862.4, + "valid_targets_min": 20307 + }, + { + "epoch": 2.6200640341515475, + "grad_norm": 0.578946446438132, + "learning_rate": 0.00017013206553045268, + "loss": 0.4499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45322614908218384, + "step": 2455, + "valid_targets_mean": 29086.3, + "valid_targets_min": 23229 + }, + { + "epoch": 2.625400213447172, + "grad_norm": 0.5302147435478608, + "learning_rate": 0.00016996589041284546, + "loss": 0.4485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4469335079193115, + "step": 2460, + "valid_targets_mean": 28731.3, + "valid_targets_min": 18282 + }, + { + "epoch": 2.6307363927427962, + "grad_norm": 0.5319875687319359, + "learning_rate": 0.00016979933595121177, + "loss": 0.4473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44532909989356995, + "step": 2465, + "valid_targets_mean": 28779.5, + "valid_targets_min": 19670 + }, + { + "epoch": 2.6360725720384206, + "grad_norm": 0.5879657232332226, + "learning_rate": 0.00016963240304858362, + "loss": 0.4469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45386573672294617, + "step": 2470, + "valid_targets_mean": 28739.6, + "valid_targets_min": 19311 + }, + { + "epoch": 2.641408751334045, + "grad_norm": 0.5768699609842823, + "learning_rate": 0.00016946509261004495, + "loss": 0.4497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4479980766773224, + "step": 2475, + "valid_targets_mean": 28779.4, + "valid_targets_min": 20019 + }, + { + "epoch": 2.6467449306296693, + "grad_norm": 0.4828563085880993, + "learning_rate": 0.00016929740554272646, + "loss": 0.4473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4485243558883667, + "step": 2480, + "valid_targets_mean": 28762.2, + "valid_targets_min": 20567 + }, + { + "epoch": 2.6520811099252937, + "grad_norm": 0.5224380520811517, + "learning_rate": 0.0001691293427558011, + "loss": 0.4499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4490726888179779, + "step": 2485, + "valid_targets_mean": 28755.4, + "valid_targets_min": 21414 + }, + { + "epoch": 2.657417289220918, + "grad_norm": 0.40870295711277854, + "learning_rate": 0.00016896090516047872, + "loss": 0.4494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.451454758644104, + "step": 2490, + "valid_targets_mean": 28876.6, + "valid_targets_min": 17603 + }, + { + "epoch": 2.6627534685165424, + "grad_norm": 0.43593658645048067, + "learning_rate": 0.0001687920936700015, + "loss": 0.4482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4449622631072998, + "step": 2495, + "valid_targets_mean": 28777.7, + "valid_targets_min": 22670 + }, + { + "epoch": 2.6680896478121667, + "grad_norm": 0.4954786841422757, + "learning_rate": 0.00016862290919963862, + "loss": 0.4482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4446685016155243, + "step": 2500, + "valid_targets_mean": 28948.7, + "valid_targets_min": 23387 + }, + { + "epoch": 2.673425827107791, + "grad_norm": 0.6120561515411104, + "learning_rate": 0.00016845335266668166, + "loss": 0.447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4459884464740753, + "step": 2505, + "valid_targets_mean": 28897.3, + "valid_targets_min": 22086 + }, + { + "epoch": 2.678762006403415, + "grad_norm": 0.48711779527463955, + "learning_rate": 0.0001682834249904394, + "loss": 0.4482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4462966322898865, + "step": 2510, + "valid_targets_mean": 28739.7, + "valid_targets_min": 20030 + }, + { + "epoch": 2.6840981856990394, + "grad_norm": 0.5285800413474548, + "learning_rate": 0.00016811312709223293, + "loss": 0.4484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4494142532348633, + "step": 2515, + "valid_targets_mean": 28892.6, + "valid_targets_min": 22882 + }, + { + "epoch": 2.6894343649946637, + "grad_norm": 0.547280626907796, + "learning_rate": 0.00016794245989539054, + "loss": 0.448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44688770174980164, + "step": 2520, + "valid_targets_mean": 28904.5, + "valid_targets_min": 22965 + }, + { + "epoch": 2.694770544290288, + "grad_norm": 0.48172505653099523, + "learning_rate": 0.00016777142432524293, + "loss": 0.4457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44005584716796875, + "step": 2525, + "valid_targets_mean": 28885.4, + "valid_targets_min": 19314 + }, + { + "epoch": 2.7001067235859124, + "grad_norm": 0.6083329167340638, + "learning_rate": 0.00016760002130911797, + "loss": 0.4472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4509948492050171, + "step": 2530, + "valid_targets_mean": 28831.2, + "valid_targets_min": 18138 + }, + { + "epoch": 2.7054429028815368, + "grad_norm": 0.49608367234784617, + "learning_rate": 0.0001674282517763358, + "loss": 0.4464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44667625427246094, + "step": 2535, + "valid_targets_mean": 28749.2, + "valid_targets_min": 20359 + }, + { + "epoch": 2.710779082177161, + "grad_norm": 0.5036192923672199, + "learning_rate": 0.0001672561166582037, + "loss": 0.4464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4464219808578491, + "step": 2540, + "valid_targets_mean": 29040.9, + "valid_targets_min": 21940 + }, + { + "epoch": 2.7161152614727855, + "grad_norm": 0.5175847995127293, + "learning_rate": 0.0001670836168880112, + "loss": 0.4474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44696956872940063, + "step": 2545, + "valid_targets_mean": 28739.1, + "valid_targets_min": 21583 + }, + { + "epoch": 2.72145144076841, + "grad_norm": 0.5467711736200832, + "learning_rate": 0.00016691075340102485, + "loss": 0.4451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4472172260284424, + "step": 2550, + "valid_targets_mean": 28817.8, + "valid_targets_min": 22683 + }, + { + "epoch": 2.726787620064034, + "grad_norm": 0.365966177831252, + "learning_rate": 0.00016673752713448314, + "loss": 0.448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44385796785354614, + "step": 2555, + "valid_targets_mean": 28766.2, + "valid_targets_min": 20755 + }, + { + "epoch": 2.7321237993596585, + "grad_norm": 0.616521778989443, + "learning_rate": 0.00016656393902759166, + "loss": 0.4475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4502711296081543, + "step": 2560, + "valid_targets_mean": 28900.8, + "valid_targets_min": 21489 + }, + { + "epoch": 2.737459978655283, + "grad_norm": 0.5919171429734628, + "learning_rate": 0.00016638999002151775, + "loss": 0.447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4452345073223114, + "step": 2565, + "valid_targets_mean": 28944.9, + "valid_targets_min": 18314 + }, + { + "epoch": 2.7427961579509073, + "grad_norm": 0.49211930960693656, + "learning_rate": 0.00016621568105938548, + "loss": 0.4464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44687044620513916, + "step": 2570, + "valid_targets_mean": 28789.1, + "valid_targets_min": 21046 + }, + { + "epoch": 2.7481323372465316, + "grad_norm": 0.4225749650195122, + "learning_rate": 0.00016604101308627053, + "loss": 0.4465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4477379322052002, + "step": 2575, + "valid_targets_mean": 28817.5, + "valid_targets_min": 20663 + }, + { + "epoch": 2.753468516542156, + "grad_norm": 0.5587518323984962, + "learning_rate": 0.00016586598704919516, + "loss": 0.4476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4498264789581299, + "step": 2580, + "valid_targets_mean": 28788.0, + "valid_targets_min": 21730 + }, + { + "epoch": 2.75880469583778, + "grad_norm": 0.5402153975521578, + "learning_rate": 0.0001656906038971229, + "loss": 0.4462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44426554441452026, + "step": 2585, + "valid_targets_mean": 28842.4, + "valid_targets_min": 20872 + }, + { + "epoch": 2.7641408751334042, + "grad_norm": 0.5402476426313789, + "learning_rate": 0.0001655148645809536, + "loss": 0.4447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4427042007446289, + "step": 2590, + "valid_targets_mean": 28991.9, + "valid_targets_min": 22105 + }, + { + "epoch": 2.7694770544290286, + "grad_norm": 0.48279597780594186, + "learning_rate": 0.0001653387700535181, + "loss": 0.4458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4428751468658447, + "step": 2595, + "valid_targets_mean": 28902.3, + "valid_targets_min": 20062 + }, + { + "epoch": 2.774813233724653, + "grad_norm": 0.4834282171961409, + "learning_rate": 0.00016516232126957312, + "loss": 0.445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44715070724487305, + "step": 2600, + "valid_targets_mean": 29066.2, + "valid_targets_min": 18312 + }, + { + "epoch": 2.7801494130202773, + "grad_norm": 0.4401907898711959, + "learning_rate": 0.00016498551918579622, + "loss": 0.4464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44535985589027405, + "step": 2605, + "valid_targets_mean": 28839.6, + "valid_targets_min": 17353 + }, + { + "epoch": 2.7854855923159016, + "grad_norm": 0.5304002176279604, + "learning_rate": 0.00016480836476078038, + "loss": 0.4466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44503068923950195, + "step": 2610, + "valid_targets_mean": 28926.2, + "valid_targets_min": 22986 + }, + { + "epoch": 2.790821771611526, + "grad_norm": 0.41619534514956896, + "learning_rate": 0.00016463085895502893, + "loss": 0.449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4478036165237427, + "step": 2615, + "valid_targets_mean": 28903.5, + "valid_targets_min": 21244 + }, + { + "epoch": 2.7961579509071504, + "grad_norm": 0.5269289182637616, + "learning_rate": 0.00016445300273095048, + "loss": 0.448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44885891675949097, + "step": 2620, + "valid_targets_mean": 28840.6, + "valid_targets_min": 22507 + }, + { + "epoch": 2.8014941302027747, + "grad_norm": 0.599103907176324, + "learning_rate": 0.0001642747970528533, + "loss": 0.4457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4468061327934265, + "step": 2625, + "valid_targets_mean": 28698.8, + "valid_targets_min": 20942 + }, + { + "epoch": 2.806830309498399, + "grad_norm": 0.4800140872410769, + "learning_rate": 0.00016409624288694056, + "loss": 0.4472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4467303156852722, + "step": 2630, + "valid_targets_mean": 28816.9, + "valid_targets_min": 18328 + }, + { + "epoch": 2.8121664887940234, + "grad_norm": 0.5266127538960025, + "learning_rate": 0.00016391734120130478, + "loss": 0.4463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4444383382797241, + "step": 2635, + "valid_targets_mean": 28910.7, + "valid_targets_min": 18160 + }, + { + "epoch": 2.8175026680896478, + "grad_norm": 0.43419276028585196, + "learning_rate": 0.00016373809296592266, + "loss": 0.4447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.443906307220459, + "step": 2640, + "valid_targets_mean": 28914.4, + "valid_targets_min": 23648 + }, + { + "epoch": 2.822838847385272, + "grad_norm": 0.491944631368096, + "learning_rate": 0.00016355849915264989, + "loss": 0.449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44785192608833313, + "step": 2645, + "valid_targets_mean": 28810.1, + "valid_targets_min": 21924 + }, + { + "epoch": 2.8281750266808965, + "grad_norm": 0.5124190278407998, + "learning_rate": 0.00016337856073521585, + "loss": 0.4467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44588106870651245, + "step": 2650, + "valid_targets_mean": 28823.3, + "valid_targets_min": 21076 + }, + { + "epoch": 2.833511205976521, + "grad_norm": 0.46000176515786123, + "learning_rate": 0.00016319827868921822, + "loss": 0.4449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44242948293685913, + "step": 2655, + "valid_targets_mean": 28862.9, + "valid_targets_min": 17089 + }, + { + "epoch": 2.838847385272145, + "grad_norm": 0.5123830710533339, + "learning_rate": 0.0001630176539921178, + "loss": 0.4467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4461979866027832, + "step": 2660, + "valid_targets_mean": 28735.5, + "valid_targets_min": 22873 + }, + { + "epoch": 2.8441835645677696, + "grad_norm": 0.5125682124858327, + "learning_rate": 0.0001628366876232333, + "loss": 0.4439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4459686875343323, + "step": 2665, + "valid_targets_mean": 28800.9, + "valid_targets_min": 21201 + }, + { + "epoch": 2.849519743863394, + "grad_norm": 0.48497008568172956, + "learning_rate": 0.00016265538056373582, + "loss": 0.4465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44951844215393066, + "step": 2670, + "valid_targets_mean": 28769.1, + "valid_targets_min": 20652 + }, + { + "epoch": 2.8548559231590183, + "grad_norm": 0.5092773446282002, + "learning_rate": 0.00016247373379664362, + "loss": 0.4457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44676268100738525, + "step": 2675, + "valid_targets_mean": 28809.3, + "valid_targets_min": 19575 + }, + { + "epoch": 2.8601921024546426, + "grad_norm": 0.4780353311067859, + "learning_rate": 0.00016229174830681683, + "loss": 0.4445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44372624158859253, + "step": 2680, + "valid_targets_mean": 28971.6, + "valid_targets_min": 15604 + }, + { + "epoch": 2.865528281750267, + "grad_norm": 0.5115745909841504, + "learning_rate": 0.00016210942508095212, + "loss": 0.4426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44047439098358154, + "step": 2685, + "valid_targets_mean": 28909.9, + "valid_targets_min": 23261 + }, + { + "epoch": 2.8708644610458913, + "grad_norm": 0.5193392457381629, + "learning_rate": 0.00016192676510757718, + "loss": 0.4457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44597327709198, + "step": 2690, + "valid_targets_mean": 28944.7, + "valid_targets_min": 21576 + }, + { + "epoch": 2.8762006403415157, + "grad_norm": 0.5480375004719434, + "learning_rate": 0.0001617437693770457, + "loss": 0.4438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4467953145503998, + "step": 2695, + "valid_targets_mean": 28884.0, + "valid_targets_min": 22438 + }, + { + "epoch": 2.88153681963714, + "grad_norm": 0.44906841653427015, + "learning_rate": 0.00016156043888153164, + "loss": 0.4441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4425533413887024, + "step": 2700, + "valid_targets_mean": 28806.5, + "valid_targets_min": 19526 + }, + { + "epoch": 2.8868729989327644, + "grad_norm": 0.621668370814208, + "learning_rate": 0.00016137677461502403, + "loss": 0.4446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4372178316116333, + "step": 2705, + "valid_targets_mean": 28990.3, + "valid_targets_min": 23225 + }, + { + "epoch": 2.8922091782283887, + "grad_norm": 0.6787756827283434, + "learning_rate": 0.00016119277757332157, + "loss": 0.443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4457907974720001, + "step": 2710, + "valid_targets_mean": 28859.2, + "valid_targets_min": 20131 + }, + { + "epoch": 2.8975453575240127, + "grad_norm": 0.45377658740918214, + "learning_rate": 0.0001610084487540272, + "loss": 0.4449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44517552852630615, + "step": 2715, + "valid_targets_mean": 28863.3, + "valid_targets_min": 22475 + }, + { + "epoch": 2.902881536819637, + "grad_norm": 0.5890326523356322, + "learning_rate": 0.00016082378915654275, + "loss": 0.4464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4380131959915161, + "step": 2720, + "valid_targets_mean": 28753.4, + "valid_targets_min": 20177 + }, + { + "epoch": 2.9082177161152614, + "grad_norm": 0.4777061022556069, + "learning_rate": 0.0001606387997820634, + "loss": 0.4458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4443918466567993, + "step": 2725, + "valid_targets_mean": 29008.7, + "valid_targets_min": 22739 + }, + { + "epoch": 2.9135538954108857, + "grad_norm": 0.4309800478585167, + "learning_rate": 0.0001604534816335724, + "loss": 0.4464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4474114179611206, + "step": 2730, + "valid_targets_mean": 28760.2, + "valid_targets_min": 22612 + }, + { + "epoch": 2.91889007470651, + "grad_norm": 0.5449182954064392, + "learning_rate": 0.0001602678357158355, + "loss": 0.4441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44207650423049927, + "step": 2735, + "valid_targets_mean": 28849.7, + "valid_targets_min": 21517 + }, + { + "epoch": 2.9242262540021344, + "grad_norm": 0.4298551008009081, + "learning_rate": 0.00016008186303539566, + "loss": 0.4438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44367218017578125, + "step": 2740, + "valid_targets_mean": 28961.5, + "valid_targets_min": 23110 + }, + { + "epoch": 2.929562433297759, + "grad_norm": 0.4524289912205744, + "learning_rate": 0.00015989556460056735, + "loss": 0.4432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4443000257015228, + "step": 2745, + "valid_targets_mean": 28859.4, + "valid_targets_min": 20708 + }, + { + "epoch": 2.934898612593383, + "grad_norm": 0.44109632853564523, + "learning_rate": 0.00015970894142143137, + "loss": 0.4423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44295409321784973, + "step": 2750, + "valid_targets_mean": 28640.2, + "valid_targets_min": 18731 + }, + { + "epoch": 2.9402347918890075, + "grad_norm": 0.6280674197563827, + "learning_rate": 0.00015952199450982915, + "loss": 0.4408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4429868459701538, + "step": 2755, + "valid_targets_mean": 28917.0, + "valid_targets_min": 21891 + }, + { + "epoch": 2.945570971184632, + "grad_norm": 0.46508843136826117, + "learning_rate": 0.0001593347248793574, + "loss": 0.4448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4444461166858673, + "step": 2760, + "valid_targets_mean": 28888.2, + "valid_targets_min": 22959 + }, + { + "epoch": 2.950907150480256, + "grad_norm": 0.46799530110116033, + "learning_rate": 0.00015914713354536243, + "loss": 0.4448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44128793478012085, + "step": 2765, + "valid_targets_mean": 28796.6, + "valid_targets_min": 21241 + }, + { + "epoch": 2.9562433297758806, + "grad_norm": 0.47772804494469634, + "learning_rate": 0.000158959221524935, + "loss": 0.4438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44201916456222534, + "step": 2770, + "valid_targets_mean": 28826.8, + "valid_targets_min": 22051 + }, + { + "epoch": 2.961579509071505, + "grad_norm": 0.44155331091775607, + "learning_rate": 0.00015877098983690442, + "loss": 0.4432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4451548755168915, + "step": 2775, + "valid_targets_mean": 28904.8, + "valid_targets_min": 20706 + }, + { + "epoch": 2.9669156883671293, + "grad_norm": 0.48382287565622245, + "learning_rate": 0.00015858243950183317, + "loss": 0.4438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4427657723426819, + "step": 2780, + "valid_targets_mean": 28867.6, + "valid_targets_min": 17286 + }, + { + "epoch": 2.9722518676627536, + "grad_norm": 0.5766134569712373, + "learning_rate": 0.0001583935715420115, + "loss": 0.4437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4461662471294403, + "step": 2785, + "valid_targets_mean": 28908.2, + "valid_targets_min": 21011 + }, + { + "epoch": 2.9775880469583775, + "grad_norm": 0.4695408328680145, + "learning_rate": 0.0001582043869814517, + "loss": 0.4424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44310444593429565, + "step": 2790, + "valid_targets_mean": 28761.7, + "valid_targets_min": 21496 + }, + { + "epoch": 2.982924226254002, + "grad_norm": 0.5000132923426082, + "learning_rate": 0.0001580148868458826, + "loss": 0.4418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44318729639053345, + "step": 2795, + "valid_targets_mean": 28784.2, + "valid_targets_min": 21259 + }, + { + "epoch": 2.9882604055496262, + "grad_norm": 0.4274040097496733, + "learning_rate": 0.0001578250721627441, + "loss": 0.4434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4456891417503357, + "step": 2800, + "valid_targets_mean": 28815.1, + "valid_targets_min": 17412 + }, + { + "epoch": 2.9935965848452506, + "grad_norm": 0.5125046838708746, + "learning_rate": 0.00015763494396118144, + "loss": 0.4423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44653642177581787, + "step": 2805, + "valid_targets_mean": 28916.4, + "valid_targets_min": 22867 + }, + { + "epoch": 2.998932764140875, + "grad_norm": 0.4815289265414247, + "learning_rate": 0.00015744450327203976, + "loss": 0.4427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4451421797275543, + "step": 2810, + "valid_targets_mean": 28983.4, + "valid_targets_min": 19654 + }, + { + "epoch": 3.0042689434364993, + "grad_norm": 0.4490098209846971, + "learning_rate": 0.00015725375112785845, + "loss": 0.4358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4360910654067993, + "step": 2815, + "valid_targets_mean": 28692.1, + "valid_targets_min": 21197 + }, + { + "epoch": 3.0096051227321237, + "grad_norm": 0.4846583619603035, + "learning_rate": 0.0001570626885628656, + "loss": 0.4375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4442782402038574, + "step": 2820, + "valid_targets_mean": 28761.2, + "valid_targets_min": 21455 + }, + { + "epoch": 3.014941302027748, + "grad_norm": 0.5843775640172333, + "learning_rate": 0.00015687131661297227, + "loss": 0.439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44055095314979553, + "step": 2825, + "valid_targets_mean": 28790.7, + "valid_targets_min": 22615 + }, + { + "epoch": 3.0202774813233724, + "grad_norm": 0.42977687830721756, + "learning_rate": 0.00015667963631576704, + "loss": 0.4377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44009584188461304, + "step": 2830, + "valid_targets_mean": 28764.4, + "valid_targets_min": 20275 + }, + { + "epoch": 3.0256136606189967, + "grad_norm": 0.5515477501528725, + "learning_rate": 0.00015648764871051022, + "loss": 0.4372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44307464361190796, + "step": 2835, + "valid_targets_mean": 28895.8, + "valid_targets_min": 22290 + }, + { + "epoch": 3.030949839914621, + "grad_norm": 0.5626533008663859, + "learning_rate": 0.00015629535483812834, + "loss": 0.4374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44284501671791077, + "step": 2840, + "valid_targets_mean": 29000.4, + "valid_targets_min": 22498 + }, + { + "epoch": 3.0362860192102454, + "grad_norm": 0.5902662848270563, + "learning_rate": 0.00015610275574120847, + "loss": 0.437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4383416771888733, + "step": 2845, + "valid_targets_mean": 28894.3, + "valid_targets_min": 20728 + }, + { + "epoch": 3.04162219850587, + "grad_norm": 0.4776689992581966, + "learning_rate": 0.00015590985246399254, + "loss": 0.4385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4417439103126526, + "step": 2850, + "valid_targets_mean": 28637.8, + "valid_targets_min": 22253 + }, + { + "epoch": 3.046958377801494, + "grad_norm": 0.5814793838248311, + "learning_rate": 0.0001557166460523717, + "loss": 0.4392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4406072497367859, + "step": 2855, + "valid_targets_mean": 28745.9, + "valid_targets_min": 21289 + }, + { + "epoch": 3.0522945570971185, + "grad_norm": 0.43664177754927597, + "learning_rate": 0.00015552313755388072, + "loss": 0.4407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43565136194229126, + "step": 2860, + "valid_targets_mean": 28927.1, + "valid_targets_min": 23185 + }, + { + "epoch": 3.057630736392743, + "grad_norm": 0.4565338949278503, + "learning_rate": 0.00015532932801769203, + "loss": 0.4384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44098299741744995, + "step": 2865, + "valid_targets_mean": 28965.3, + "valid_targets_min": 19321 + }, + { + "epoch": 3.062966915688367, + "grad_norm": 0.5857395045348511, + "learning_rate": 0.0001551352184946105, + "loss": 0.4374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43504267930984497, + "step": 2870, + "valid_targets_mean": 28885.3, + "valid_targets_min": 22399 + }, + { + "epoch": 3.0683030949839916, + "grad_norm": 0.44489437983437763, + "learning_rate": 0.00015494081003706732, + "loss": 0.439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4336409568786621, + "step": 2875, + "valid_targets_mean": 28808.5, + "valid_targets_min": 19901 + }, + { + "epoch": 3.073639274279616, + "grad_norm": 0.5079821686131969, + "learning_rate": 0.00015474610369911447, + "loss": 0.4359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43570592999458313, + "step": 2880, + "valid_targets_mean": 28721.2, + "valid_targets_min": 20846 + }, + { + "epoch": 3.0789754535752403, + "grad_norm": 0.5466811796972372, + "learning_rate": 0.00015455110053641904, + "loss": 0.4365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43693655729293823, + "step": 2885, + "valid_targets_mean": 28707.8, + "valid_targets_min": 22286 + }, + { + "epoch": 3.0843116328708646, + "grad_norm": 0.4747809130741953, + "learning_rate": 0.00015435580160625746, + "loss": 0.4361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4353131651878357, + "step": 2890, + "valid_targets_mean": 28815.9, + "valid_targets_min": 19218 + }, + { + "epoch": 3.089647812166489, + "grad_norm": 0.5577315428315981, + "learning_rate": 0.00015416020796750964, + "loss": 0.4375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4367016553878784, + "step": 2895, + "valid_targets_mean": 28818.8, + "valid_targets_min": 18724 + }, + { + "epoch": 3.0949839914621133, + "grad_norm": 0.4377505907429175, + "learning_rate": 0.00015396432068065356, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43655478954315186, + "step": 2900, + "valid_targets_mean": 28768.1, + "valid_targets_min": 17581 + }, + { + "epoch": 3.1003201707577372, + "grad_norm": 0.5422717618043638, + "learning_rate": 0.00015376814080775915, + "loss": 0.4391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43886345624923706, + "step": 2905, + "valid_targets_mean": 28785.8, + "valid_targets_min": 20545 + }, + { + "epoch": 3.1056563500533616, + "grad_norm": 0.4737668348911352, + "learning_rate": 0.00015357166941248277, + "loss": 0.4376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43835288286209106, + "step": 2910, + "valid_targets_mean": 28717.6, + "valid_targets_min": 22895 + }, + { + "epoch": 3.110992529348986, + "grad_norm": 0.543871526659854, + "learning_rate": 0.00015337490756006132, + "loss": 0.4381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4373202919960022, + "step": 2915, + "valid_targets_mean": 28874.1, + "valid_targets_min": 21424 + }, + { + "epoch": 3.1163287086446103, + "grad_norm": 0.45527342072667687, + "learning_rate": 0.00015317785631730661, + "loss": 0.4389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.440227746963501, + "step": 2920, + "valid_targets_mean": 28739.2, + "valid_targets_min": 21858 + }, + { + "epoch": 3.1216648879402347, + "grad_norm": 0.49916830905132054, + "learning_rate": 0.00015298051675259933, + "loss": 0.4376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43711093068122864, + "step": 2925, + "valid_targets_mean": 28903.5, + "valid_targets_min": 19999 + }, + { + "epoch": 3.127001067235859, + "grad_norm": 0.427372846116701, + "learning_rate": 0.00015278288993588346, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4339505434036255, + "step": 2930, + "valid_targets_mean": 28693.9, + "valid_targets_min": 17885 + }, + { + "epoch": 3.1323372465314834, + "grad_norm": 0.35038643196964536, + "learning_rate": 0.00015258497693866057, + "loss": 0.4351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43530455231666565, + "step": 2935, + "valid_targets_mean": 28755.8, + "valid_targets_min": 22373 + }, + { + "epoch": 3.1376734258271077, + "grad_norm": 0.4440862785337304, + "learning_rate": 0.00015238677883398362, + "loss": 0.4376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43910306692123413, + "step": 2940, + "valid_targets_mean": 28886.5, + "valid_targets_min": 21548 + }, + { + "epoch": 3.143009605122732, + "grad_norm": 0.5621426459146843, + "learning_rate": 0.0001521882966964515, + "loss": 0.4392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43817323446273804, + "step": 2945, + "valid_targets_mean": 28765.8, + "valid_targets_min": 16515 + }, + { + "epoch": 3.1483457844183564, + "grad_norm": 0.5053752189040438, + "learning_rate": 0.00015198953160220312, + "loss": 0.4354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4326125383377075, + "step": 2950, + "valid_targets_mean": 28974.8, + "valid_targets_min": 17895 + }, + { + "epoch": 3.153681963713981, + "grad_norm": 0.47984183388372986, + "learning_rate": 0.00015179048462891145, + "loss": 0.4385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4363112449645996, + "step": 2955, + "valid_targets_mean": 29044.7, + "valid_targets_min": 23385 + }, + { + "epoch": 3.159018143009605, + "grad_norm": 0.5107486500128496, + "learning_rate": 0.00015159115685577774, + "loss": 0.4353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43246275186538696, + "step": 2960, + "valid_targets_mean": 28840.2, + "valid_targets_min": 22072 + }, + { + "epoch": 3.1643543223052295, + "grad_norm": 0.4836933274092199, + "learning_rate": 0.00015139154936352585, + "loss": 0.4346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43785497546195984, + "step": 2965, + "valid_targets_mean": 28919.9, + "valid_targets_min": 22145 + }, + { + "epoch": 3.169690501600854, + "grad_norm": 0.5543478553177675, + "learning_rate": 0.00015119166323439604, + "loss": 0.4351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4362708330154419, + "step": 2970, + "valid_targets_mean": 28837.1, + "valid_targets_min": 21022 + }, + { + "epoch": 3.175026680896478, + "grad_norm": 0.5400264191824763, + "learning_rate": 0.00015099149955213943, + "loss": 0.438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4363650977611542, + "step": 2975, + "valid_targets_mean": 28916.7, + "valid_targets_min": 23047 + }, + { + "epoch": 3.1803628601921026, + "grad_norm": 0.568218023832969, + "learning_rate": 0.00015079105940201198, + "loss": 0.4362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4391043782234192, + "step": 2980, + "valid_targets_mean": 29012.3, + "valid_targets_min": 21232 + }, + { + "epoch": 3.185699039487727, + "grad_norm": 0.531139321909535, + "learning_rate": 0.00015059034387076854, + "loss": 0.4393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4397522211074829, + "step": 2985, + "valid_targets_mean": 28765.6, + "valid_targets_min": 22435 + }, + { + "epoch": 3.1910352187833513, + "grad_norm": 0.4609349250065392, + "learning_rate": 0.00015038935404665705, + "loss": 0.4357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4346761703491211, + "step": 2990, + "valid_targets_mean": 28826.0, + "valid_targets_min": 21307 + }, + { + "epoch": 3.1963713980789756, + "grad_norm": 0.4000357269001568, + "learning_rate": 0.00015018809101941268, + "loss": 0.4376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43649059534072876, + "step": 2995, + "valid_targets_mean": 28806.0, + "valid_targets_min": 21706 + }, + { + "epoch": 3.2017075773746, + "grad_norm": 0.37697317071252967, + "learning_rate": 0.0001499865558802518, + "loss": 0.4366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43681079149246216, + "step": 3000, + "valid_targets_mean": 28887.8, + "valid_targets_min": 21593 + }, + { + "epoch": 3.207043756670224, + "grad_norm": 0.46675713799883556, + "learning_rate": 0.00014978474972186621, + "loss": 0.436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4359115958213806, + "step": 3005, + "valid_targets_mean": 28906.7, + "valid_targets_min": 21243 + }, + { + "epoch": 3.2123799359658483, + "grad_norm": 0.4794146972351641, + "learning_rate": 0.00014958267363841707, + "loss": 0.4357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43594199419021606, + "step": 3010, + "valid_targets_mean": 28788.4, + "valid_targets_min": 21892 + }, + { + "epoch": 3.2177161152614726, + "grad_norm": 0.511084600739757, + "learning_rate": 0.00014938032872552904, + "loss": 0.4409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4422481954097748, + "step": 3015, + "valid_targets_mean": 28903.2, + "valid_targets_min": 22725 + }, + { + "epoch": 3.223052294557097, + "grad_norm": 0.5081539910592223, + "learning_rate": 0.0001491777160802843, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4305049777030945, + "step": 3020, + "valid_targets_mean": 28737.1, + "valid_targets_min": 17466 + }, + { + "epoch": 3.2283884738527213, + "grad_norm": 0.4828869142114176, + "learning_rate": 0.00014897483680121669, + "loss": 0.435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43581363558769226, + "step": 3025, + "valid_targets_mean": 28965.7, + "valid_targets_min": 20422 + }, + { + "epoch": 3.2337246531483457, + "grad_norm": 0.4578951471605822, + "learning_rate": 0.0001487716919883057, + "loss": 0.4368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43529069423675537, + "step": 3030, + "valid_targets_mean": 28887.0, + "valid_targets_min": 21186 + }, + { + "epoch": 3.23906083244397, + "grad_norm": 0.5472268575085925, + "learning_rate": 0.00014856828274297044, + "loss": 0.4377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4369294345378876, + "step": 3035, + "valid_targets_mean": 28764.9, + "valid_targets_min": 18549 + }, + { + "epoch": 3.2443970117395944, + "grad_norm": 0.48058959360049996, + "learning_rate": 0.00014836461016806378, + "loss": 0.4378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43576347827911377, + "step": 3040, + "valid_targets_mean": 28918.6, + "valid_targets_min": 22352 + }, + { + "epoch": 3.2497331910352187, + "grad_norm": 0.47631716841746413, + "learning_rate": 0.0001481606753678663, + "loss": 0.436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43823981285095215, + "step": 3045, + "valid_targets_mean": 28839.2, + "valid_targets_min": 21158 + }, + { + "epoch": 3.255069370330843, + "grad_norm": 0.4644880270827397, + "learning_rate": 0.00014795647944808036, + "loss": 0.4364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43759581446647644, + "step": 3050, + "valid_targets_mean": 28845.9, + "valid_targets_min": 22918 + }, + { + "epoch": 3.2604055496264674, + "grad_norm": 0.5072907213602093, + "learning_rate": 0.000147752023515824, + "loss": 0.4351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4352570176124573, + "step": 3055, + "valid_targets_mean": 28938.5, + "valid_targets_min": 20322 + }, + { + "epoch": 3.265741728922092, + "grad_norm": 0.507176783487205, + "learning_rate": 0.00014754730867962506, + "loss": 0.4365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43387681245803833, + "step": 3060, + "valid_targets_mean": 28740.9, + "valid_targets_min": 21035 + }, + { + "epoch": 3.271077908217716, + "grad_norm": 0.48370474732246943, + "learning_rate": 0.00014734233604941517, + "loss": 0.4363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43365657329559326, + "step": 3065, + "valid_targets_mean": 28687.2, + "valid_targets_min": 20895 + }, + { + "epoch": 3.2764140875133405, + "grad_norm": 0.46332555017627197, + "learning_rate": 0.00014713710673652355, + "loss": 0.4328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4362318217754364, + "step": 3070, + "valid_targets_mean": 28952.4, + "valid_targets_min": 22099 + }, + { + "epoch": 3.281750266808965, + "grad_norm": 0.48378410659769616, + "learning_rate": 0.00014693162185367114, + "loss": 0.4368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43846186995506287, + "step": 3075, + "valid_targets_mean": 28972.7, + "valid_targets_min": 18619 + }, + { + "epoch": 3.287086446104589, + "grad_norm": 0.561520367354519, + "learning_rate": 0.00014672588251496467, + "loss": 0.4355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43675142526626587, + "step": 3080, + "valid_targets_mean": 28849.4, + "valid_targets_min": 17793 + }, + { + "epoch": 3.2924226254002136, + "grad_norm": 0.44156610019257575, + "learning_rate": 0.00014651988983589034, + "loss": 0.433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4352341294288635, + "step": 3085, + "valid_targets_mean": 28802.1, + "valid_targets_min": 21724 + }, + { + "epoch": 3.297758804695838, + "grad_norm": 0.4197577896293227, + "learning_rate": 0.00014631364493330798, + "loss": 0.4348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43272507190704346, + "step": 3090, + "valid_targets_mean": 28770.6, + "valid_targets_min": 21001 + }, + { + "epoch": 3.3030949839914623, + "grad_norm": 0.4886419166778624, + "learning_rate": 0.00014610714892544492, + "loss": 0.4348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4325237572193146, + "step": 3095, + "valid_targets_mean": 28874.4, + "valid_targets_min": 19509 + }, + { + "epoch": 3.3084311632870866, + "grad_norm": 0.459899265230845, + "learning_rate": 0.00014590040293189, + "loss": 0.4364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4408683180809021, + "step": 3100, + "valid_targets_mean": 28998.2, + "valid_targets_min": 22881 + }, + { + "epoch": 3.313767342582711, + "grad_norm": 0.5373706447277511, + "learning_rate": 0.0001456934080735874, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43423405289649963, + "step": 3105, + "valid_targets_mean": 28955.8, + "valid_targets_min": 22814 + }, + { + "epoch": 3.3191035218783353, + "grad_norm": 0.4593122288739941, + "learning_rate": 0.00014548616547283057, + "loss": 0.436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4395608603954315, + "step": 3110, + "valid_targets_mean": 28559.3, + "valid_targets_min": 20401 + }, + { + "epoch": 3.3244397011739593, + "grad_norm": 0.5534304935891788, + "learning_rate": 0.0001452786762532562, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43570777773857117, + "step": 3115, + "valid_targets_mean": 28819.2, + "valid_targets_min": 17157 + }, + { + "epoch": 3.3297758804695836, + "grad_norm": 0.5280281388839083, + "learning_rate": 0.00014507094153983817, + "loss": 0.4352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4362625777721405, + "step": 3120, + "valid_targets_mean": 28904.0, + "valid_targets_min": 22555 + }, + { + "epoch": 3.335112059765208, + "grad_norm": 0.41320446076613393, + "learning_rate": 0.00014486296245888133, + "loss": 0.434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43435296416282654, + "step": 3125, + "valid_targets_mean": 28724.5, + "valid_targets_min": 19569 + }, + { + "epoch": 3.3404482390608323, + "grad_norm": 0.48360144671687794, + "learning_rate": 0.00014465474013801547, + "loss": 0.4329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4317402243614197, + "step": 3130, + "valid_targets_mean": 28933.7, + "valid_targets_min": 19488 + }, + { + "epoch": 3.3457844183564567, + "grad_norm": 0.5029351455588194, + "learning_rate": 0.00014444627570618917, + "loss": 0.4341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4321380853652954, + "step": 3135, + "valid_targets_mean": 28885.6, + "valid_targets_min": 22972 + }, + { + "epoch": 3.351120597652081, + "grad_norm": 0.4618742374392624, + "learning_rate": 0.00014423757029366371, + "loss": 0.4338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4360083341598511, + "step": 3140, + "valid_targets_mean": 28724.2, + "valid_targets_min": 22321 + }, + { + "epoch": 3.3564567769477054, + "grad_norm": 0.500587834014976, + "learning_rate": 0.00014402862503200693, + "loss": 0.4331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4348517954349518, + "step": 3145, + "valid_targets_mean": 28856.2, + "valid_targets_min": 20609 + }, + { + "epoch": 3.3617929562433297, + "grad_norm": 0.46784044054670815, + "learning_rate": 0.00014381944105408713, + "loss": 0.4331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4310871660709381, + "step": 3150, + "valid_targets_mean": 28724.0, + "valid_targets_min": 20282 + }, + { + "epoch": 3.367129135538954, + "grad_norm": 0.4805330497133465, + "learning_rate": 0.00014361001949406683, + "loss": 0.4345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4370986819267273, + "step": 3155, + "valid_targets_mean": 28870.4, + "valid_targets_min": 22529 + }, + { + "epoch": 3.3724653148345785, + "grad_norm": 0.557962978167396, + "learning_rate": 0.00014340036148739667, + "loss": 0.4366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.433628112077713, + "step": 3160, + "valid_targets_mean": 28784.4, + "valid_targets_min": 13864 + }, + { + "epoch": 3.377801494130203, + "grad_norm": 0.47468145347567137, + "learning_rate": 0.0001431904681708094, + "loss": 0.4326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43612152338027954, + "step": 3165, + "valid_targets_mean": 28685.5, + "valid_targets_min": 22206 + }, + { + "epoch": 3.383137673425827, + "grad_norm": 0.47367311151064023, + "learning_rate": 0.0001429803406823134, + "loss": 0.4361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4309577941894531, + "step": 3170, + "valid_targets_mean": 28744.8, + "valid_targets_min": 21212 + }, + { + "epoch": 3.3884738527214515, + "grad_norm": 0.4293787714289241, + "learning_rate": 0.00014276998016118683, + "loss": 0.435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43373751640319824, + "step": 3175, + "valid_targets_mean": 28691.5, + "valid_targets_min": 17108 + }, + { + "epoch": 3.393810032017076, + "grad_norm": 0.46229360735788555, + "learning_rate": 0.0001425593877479712, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43639492988586426, + "step": 3180, + "valid_targets_mean": 28808.1, + "valid_targets_min": 21564 + }, + { + "epoch": 3.3991462113127002, + "grad_norm": 0.502035175620512, + "learning_rate": 0.0001423485645844655, + "loss": 0.434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4358476996421814, + "step": 3185, + "valid_targets_mean": 28687.7, + "valid_targets_min": 18846 + }, + { + "epoch": 3.4044823906083246, + "grad_norm": 0.40960475991145945, + "learning_rate": 0.0001421375118137196, + "loss": 0.4327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4317413568496704, + "step": 3190, + "valid_targets_mean": 28913.0, + "valid_targets_min": 23478 + }, + { + "epoch": 3.409818569903949, + "grad_norm": 0.4349549275319588, + "learning_rate": 0.00014192623058002833, + "loss": 0.4338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4319398105144501, + "step": 3195, + "valid_targets_mean": 28981.6, + "valid_targets_min": 22411 + }, + { + "epoch": 3.4151547491995733, + "grad_norm": 0.4854158214544209, + "learning_rate": 0.00014171472202892526, + "loss": 0.4329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4335978627204895, + "step": 3200, + "valid_targets_mean": 28860.3, + "valid_targets_min": 21640 + }, + { + "epoch": 3.420490928495197, + "grad_norm": 0.44519855071377373, + "learning_rate": 0.00014150298730717642, + "loss": 0.4352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43621671199798584, + "step": 3205, + "valid_targets_mean": 29066.8, + "valid_targets_min": 21390 + }, + { + "epoch": 3.4258271077908216, + "grad_norm": 0.507964411024077, + "learning_rate": 0.00014129102756277404, + "loss": 0.4344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4369572699069977, + "step": 3210, + "valid_targets_mean": 28842.6, + "valid_targets_min": 20748 + }, + { + "epoch": 3.431163287086446, + "grad_norm": 0.45705877101910725, + "learning_rate": 0.00014107884394493045, + "loss": 0.433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43259197473526, + "step": 3215, + "valid_targets_mean": 28897.4, + "valid_targets_min": 22273 + }, + { + "epoch": 3.4364994663820703, + "grad_norm": 0.5951471488936144, + "learning_rate": 0.00014086643760407184, + "loss": 0.4359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4341096580028534, + "step": 3220, + "valid_targets_mean": 28996.3, + "valid_targets_min": 18485 + }, + { + "epoch": 3.4418356456776946, + "grad_norm": 0.44496514528606296, + "learning_rate": 0.00014065380969183178, + "loss": 0.4337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4341355860233307, + "step": 3225, + "valid_targets_mean": 28753.5, + "valid_targets_min": 18329 + }, + { + "epoch": 3.447171824973319, + "grad_norm": 0.4751836858048521, + "learning_rate": 0.00014044096136104538, + "loss": 0.4328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43249812722206116, + "step": 3230, + "valid_targets_mean": 29014.2, + "valid_targets_min": 21394 + }, + { + "epoch": 3.4525080042689433, + "grad_norm": 0.4418432182242359, + "learning_rate": 0.0001402278937657427, + "loss": 0.4323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43450066447257996, + "step": 3235, + "valid_targets_mean": 28716.2, + "valid_targets_min": 20397 + }, + { + "epoch": 3.4578441835645677, + "grad_norm": 0.436845201857216, + "learning_rate": 0.0001400146080611426, + "loss": 0.4347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43189460039138794, + "step": 3240, + "valid_targets_mean": 28895.9, + "valid_targets_min": 20984 + }, + { + "epoch": 3.463180362860192, + "grad_norm": 0.49129311118631425, + "learning_rate": 0.00013980110540364664, + "loss": 0.4326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43245309591293335, + "step": 3245, + "valid_targets_mean": 28779.6, + "valid_targets_min": 18347 + }, + { + "epoch": 3.4685165421558164, + "grad_norm": 0.41406368494446794, + "learning_rate": 0.00013958738695083253, + "loss": 0.433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42854857444763184, + "step": 3250, + "valid_targets_mean": 29027.9, + "valid_targets_min": 19358 + }, + { + "epoch": 3.4738527214514408, + "grad_norm": 0.3914897379980043, + "learning_rate": 0.00013937345386144797, + "loss": 0.4322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4313948154449463, + "step": 3255, + "valid_targets_mean": 28844.7, + "valid_targets_min": 19376 + }, + { + "epoch": 3.479188900747065, + "grad_norm": 0.44171395273171776, + "learning_rate": 0.00013915930729540448, + "loss": 0.4336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43444424867630005, + "step": 3260, + "valid_targets_mean": 28860.9, + "valid_targets_min": 22790 + }, + { + "epoch": 3.4845250800426895, + "grad_norm": 0.44474343829313423, + "learning_rate": 0.00013894494841377102, + "loss": 0.4331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4332994520664215, + "step": 3265, + "valid_targets_mean": 28780.2, + "valid_targets_min": 18196 + }, + { + "epoch": 3.489861259338314, + "grad_norm": 0.49347413137837826, + "learning_rate": 0.00013873037837876758, + "loss": 0.4317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43504035472869873, + "step": 3270, + "valid_targets_mean": 28805.4, + "valid_targets_min": 19505 + }, + { + "epoch": 3.495197438633938, + "grad_norm": 0.43703657771662985, + "learning_rate": 0.0001385155983537591, + "loss": 0.4353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4368758797645569, + "step": 3275, + "valid_targets_mean": 28873.7, + "valid_targets_min": 22760 + }, + { + "epoch": 3.5005336179295625, + "grad_norm": 0.49549736446056913, + "learning_rate": 0.00013830060950324898, + "loss": 0.4326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42575693130493164, + "step": 3280, + "valid_targets_mean": 28792.0, + "valid_targets_min": 22707 + }, + { + "epoch": 3.505869797225187, + "grad_norm": 0.4582256590179247, + "learning_rate": 0.0001380854129928729, + "loss": 0.4331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4306070804595947, + "step": 3285, + "valid_targets_mean": 28942.3, + "valid_targets_min": 21332 + }, + { + "epoch": 3.5112059765208112, + "grad_norm": 0.40229320048606765, + "learning_rate": 0.0001378700099893924, + "loss": 0.4346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4349821209907532, + "step": 3290, + "valid_targets_mean": 28724.8, + "valid_targets_min": 19366 + }, + { + "epoch": 3.5165421558164356, + "grad_norm": 0.42753744321130677, + "learning_rate": 0.0001376544016606886, + "loss": 0.431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4316447973251343, + "step": 3295, + "valid_targets_mean": 28727.2, + "valid_targets_min": 19929 + }, + { + "epoch": 3.52187833511206, + "grad_norm": 0.41143315982773954, + "learning_rate": 0.00013743858917575588, + "loss": 0.4323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4331210255622864, + "step": 3300, + "valid_targets_mean": 28951.2, + "valid_targets_min": 23322 + }, + { + "epoch": 3.5272145144076843, + "grad_norm": 0.3661504334505945, + "learning_rate": 0.00013722257370469545, + "loss": 0.4316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43106207251548767, + "step": 3305, + "valid_targets_mean": 28848.0, + "valid_targets_min": 20950 + }, + { + "epoch": 3.5325506937033087, + "grad_norm": 0.3145847072932158, + "learning_rate": 0.00013700635641870923, + "loss": 0.4337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4317154288291931, + "step": 3310, + "valid_targets_mean": 28951.6, + "valid_targets_min": 22482 + }, + { + "epoch": 3.537886872998933, + "grad_norm": 0.3760485517392338, + "learning_rate": 0.00013678993849009313, + "loss": 0.4316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43668973445892334, + "step": 3315, + "valid_targets_mean": 28854.2, + "valid_targets_min": 20523 + }, + { + "epoch": 3.5432230522945574, + "grad_norm": 0.33028254222060127, + "learning_rate": 0.00013657332109223117, + "loss": 0.434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4305395781993866, + "step": 3320, + "valid_targets_mean": 28878.3, + "valid_targets_min": 21290 + }, + { + "epoch": 3.5485592315901813, + "grad_norm": 0.5105044696911828, + "learning_rate": 0.00013635650539958863, + "loss": 0.4328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4329110383987427, + "step": 3325, + "valid_targets_mean": 28852.5, + "valid_targets_min": 20940 + }, + { + "epoch": 3.5538954108858056, + "grad_norm": 0.44178193667691457, + "learning_rate": 0.00013613949258770608, + "loss": 0.4326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4327603578567505, + "step": 3330, + "valid_targets_mean": 28809.5, + "valid_targets_min": 22051 + }, + { + "epoch": 3.55923159018143, + "grad_norm": 0.4592497401974742, + "learning_rate": 0.00013592228383319273, + "loss": 0.4332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43314966559410095, + "step": 3335, + "valid_targets_mean": 28734.6, + "valid_targets_min": 21478 + }, + { + "epoch": 3.5645677694770543, + "grad_norm": 0.39984373491737396, + "learning_rate": 0.00013570488031372022, + "loss": 0.4331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4362226128578186, + "step": 3340, + "valid_targets_mean": 28880.2, + "valid_targets_min": 22299 + }, + { + "epoch": 3.5699039487726787, + "grad_norm": 0.42971496493126365, + "learning_rate": 0.00013548728320801612, + "loss": 0.4338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43295514583587646, + "step": 3345, + "valid_targets_mean": 28929.0, + "valid_targets_min": 23270 + }, + { + "epoch": 3.575240128068303, + "grad_norm": 0.46650323926697734, + "learning_rate": 0.00013526949369585764, + "loss": 0.4314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4373741149902344, + "step": 3350, + "valid_targets_mean": 29128.0, + "valid_targets_min": 22056 + }, + { + "epoch": 3.5805763073639274, + "grad_norm": 0.48225276813971074, + "learning_rate": 0.0001350515129580652, + "loss": 0.4318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43129345774650574, + "step": 3355, + "valid_targets_mean": 28706.7, + "valid_targets_min": 17828 + }, + { + "epoch": 3.5859124866595518, + "grad_norm": 0.4629045352920466, + "learning_rate": 0.000134833342176496, + "loss": 0.4337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4332507252693176, + "step": 3360, + "valid_targets_mean": 28870.7, + "valid_targets_min": 21083 + }, + { + "epoch": 3.591248665955176, + "grad_norm": 0.47206526382533237, + "learning_rate": 0.00013461498253403758, + "loss": 0.4337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4355852007865906, + "step": 3365, + "valid_targets_mean": 28735.8, + "valid_targets_min": 18046 + }, + { + "epoch": 3.5965848452508005, + "grad_norm": 0.4700787626748754, + "learning_rate": 0.00013439643521460153, + "loss": 0.43, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43339934945106506, + "step": 3370, + "valid_targets_mean": 28750.6, + "valid_targets_min": 19141 + }, + { + "epoch": 3.601921024546425, + "grad_norm": 0.43350780960940405, + "learning_rate": 0.00013417770140311693, + "loss": 0.4315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4338434338569641, + "step": 3375, + "valid_targets_mean": 28719.0, + "valid_targets_min": 19379 + }, + { + "epoch": 3.607257203842049, + "grad_norm": 0.4013004042105287, + "learning_rate": 0.000133958782285524, + "loss": 0.4304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4283779561519623, + "step": 3380, + "valid_targets_mean": 28811.4, + "valid_targets_min": 18503 + }, + { + "epoch": 3.6125933831376735, + "grad_norm": 0.5388393269800589, + "learning_rate": 0.00013373967904876775, + "loss": 0.4337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4304119050502777, + "step": 3385, + "valid_targets_mean": 29018.2, + "valid_targets_min": 21958 + }, + { + "epoch": 3.617929562433298, + "grad_norm": 0.4057390217196227, + "learning_rate": 0.00013352039288079133, + "loss": 0.4306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4319956600666046, + "step": 3390, + "valid_targets_mean": 28919.5, + "valid_targets_min": 22140 + }, + { + "epoch": 3.6232657417289222, + "grad_norm": 0.5368401785672807, + "learning_rate": 0.00013330092497052976, + "loss": 0.433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.433579683303833, + "step": 3395, + "valid_targets_mean": 28931.2, + "valid_targets_min": 19698 + }, + { + "epoch": 3.628601921024546, + "grad_norm": 0.36185761939563554, + "learning_rate": 0.00013308127650790352, + "loss": 0.4319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43094661831855774, + "step": 3400, + "valid_targets_mean": 28791.9, + "valid_targets_min": 18610 + }, + { + "epoch": 3.6339381003201705, + "grad_norm": 0.4425115168030993, + "learning_rate": 0.00013286144868381188, + "loss": 0.4325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4321828782558441, + "step": 3405, + "valid_targets_mean": 28859.5, + "valid_targets_min": 17625 + }, + { + "epoch": 3.639274279615795, + "grad_norm": 0.5914395427100793, + "learning_rate": 0.0001326414426901266, + "loss": 0.4321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4337488114833832, + "step": 3410, + "valid_targets_mean": 28888.0, + "valid_targets_min": 22467 + }, + { + "epoch": 3.644610458911419, + "grad_norm": 0.4870561175085104, + "learning_rate": 0.00013242125971968556, + "loss": 0.4321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42824339866638184, + "step": 3415, + "valid_targets_mean": 28959.8, + "valid_targets_min": 21801 + }, + { + "epoch": 3.6499466382070436, + "grad_norm": 0.44613044833777615, + "learning_rate": 0.00013220090096628616, + "loss": 0.4336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43413907289505005, + "step": 3420, + "valid_targets_mean": 28731.1, + "valid_targets_min": 21438 + }, + { + "epoch": 3.655282817502668, + "grad_norm": 0.4094552612110865, + "learning_rate": 0.00013198036762467868, + "loss": 0.432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4328404664993286, + "step": 3425, + "valid_targets_mean": 28759.0, + "valid_targets_min": 21885 + }, + { + "epoch": 3.6606189967982923, + "grad_norm": 0.4966656728086518, + "learning_rate": 0.00013175966089056025, + "loss": 0.4316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4348108172416687, + "step": 3430, + "valid_targets_mean": 28708.3, + "valid_targets_min": 22466 + }, + { + "epoch": 3.6659551760939166, + "grad_norm": 0.4369493068182657, + "learning_rate": 0.00013153878196056788, + "loss": 0.4304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4332895278930664, + "step": 3435, + "valid_targets_mean": 28939.6, + "valid_targets_min": 21070 + }, + { + "epoch": 3.671291355389541, + "grad_norm": 0.445549024199376, + "learning_rate": 0.00013131773203227238, + "loss": 0.4304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43516385555267334, + "step": 3440, + "valid_targets_mean": 28753.2, + "valid_targets_min": 22085 + }, + { + "epoch": 3.6766275346851653, + "grad_norm": 0.48475802506018867, + "learning_rate": 0.00013109651230417158, + "loss": 0.4295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42581403255462646, + "step": 3445, + "valid_targets_mean": 28848.2, + "valid_targets_min": 22898 + }, + { + "epoch": 3.6819637139807897, + "grad_norm": 0.41403337615437824, + "learning_rate": 0.000130875123975684, + "loss": 0.4321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4342615604400635, + "step": 3450, + "valid_targets_mean": 28758.9, + "valid_targets_min": 20344 + }, + { + "epoch": 3.687299893276414, + "grad_norm": 0.469971436134684, + "learning_rate": 0.00013065356824714218, + "loss": 0.4318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4311075210571289, + "step": 3455, + "valid_targets_mean": 28824.2, + "valid_targets_min": 20941 + }, + { + "epoch": 3.6926360725720384, + "grad_norm": 0.4678101527655715, + "learning_rate": 0.0001304318463197864, + "loss": 0.4303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42798611521720886, + "step": 3460, + "valid_targets_mean": 28862.6, + "valid_targets_min": 19023 + }, + { + "epoch": 3.6979722518676628, + "grad_norm": 0.408590337655623, + "learning_rate": 0.00013020995939575801, + "loss": 0.4293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4305270314216614, + "step": 3465, + "valid_targets_mean": 28937.8, + "valid_targets_min": 23337 + }, + { + "epoch": 3.703308431163287, + "grad_norm": 0.4656795617970675, + "learning_rate": 0.0001299879086780929, + "loss": 0.4286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4290810525417328, + "step": 3470, + "valid_targets_mean": 28860.1, + "valid_targets_min": 19674 + }, + { + "epoch": 3.7086446104589115, + "grad_norm": 0.5658992971599166, + "learning_rate": 0.000129765695370715, + "loss": 0.4305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4292992353439331, + "step": 3475, + "valid_targets_mean": 28804.4, + "valid_targets_min": 17714 + }, + { + "epoch": 3.713980789754536, + "grad_norm": 0.39311007426573547, + "learning_rate": 0.00012954332067842994, + "loss": 0.4308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42862972617149353, + "step": 3480, + "valid_targets_mean": 28798.6, + "valid_targets_min": 22104 + }, + { + "epoch": 3.71931696905016, + "grad_norm": 0.3542095011377809, + "learning_rate": 0.00012932078580691817, + "loss": 0.4325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4315662384033203, + "step": 3485, + "valid_targets_mean": 28961.7, + "valid_targets_min": 22930 + }, + { + "epoch": 3.7246531483457845, + "grad_norm": 0.4028606369164767, + "learning_rate": 0.00012909809196272865, + "loss": 0.4289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42671799659729004, + "step": 3490, + "valid_targets_mean": 28998.3, + "valid_targets_min": 21922 + }, + { + "epoch": 3.729989327641409, + "grad_norm": 0.44110145771370024, + "learning_rate": 0.00012887524035327233, + "loss": 0.4309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4318047761917114, + "step": 3495, + "valid_targets_mean": 28855.4, + "valid_targets_min": 23180 + }, + { + "epoch": 3.7353255069370332, + "grad_norm": 0.4445599952668899, + "learning_rate": 0.00012865223218681555, + "loss": 0.4316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4287042021751404, + "step": 3500, + "valid_targets_mean": 28774.2, + "valid_targets_min": 22317 + }, + { + "epoch": 3.7406616862326576, + "grad_norm": 0.4660901172079049, + "learning_rate": 0.0001284290686724733, + "loss": 0.4278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42891907691955566, + "step": 3505, + "valid_targets_mean": 28782.6, + "valid_targets_min": 19569 + }, + { + "epoch": 3.745997865528282, + "grad_norm": 0.39478144276395094, + "learning_rate": 0.00012820575102020308, + "loss": 0.4291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42791077494621277, + "step": 3510, + "valid_targets_mean": 28934.6, + "valid_targets_min": 21724 + }, + { + "epoch": 3.7513340448239063, + "grad_norm": 0.4498652413048794, + "learning_rate": 0.00012798228044079795, + "loss": 0.4301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4289936423301697, + "step": 3515, + "valid_targets_mean": 28785.5, + "valid_targets_min": 22540 + }, + { + "epoch": 3.7566702241195307, + "grad_norm": 0.37725205122505695, + "learning_rate": 0.0001277586581458802, + "loss": 0.4281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42873474955558777, + "step": 3520, + "valid_targets_mean": 28793.5, + "valid_targets_min": 20282 + }, + { + "epoch": 3.762006403415155, + "grad_norm": 0.3851394073217626, + "learning_rate": 0.00012753488534789458, + "loss": 0.429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4283927083015442, + "step": 3525, + "valid_targets_mean": 28736.1, + "valid_targets_min": 20171 + }, + { + "epoch": 3.767342582710779, + "grad_norm": 0.3892498423056188, + "learning_rate": 0.00012731096326010194, + "loss": 0.4314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4264993667602539, + "step": 3530, + "valid_targets_mean": 28871.7, + "valid_targets_min": 20214 + }, + { + "epoch": 3.7726787620064033, + "grad_norm": 0.37455931834224676, + "learning_rate": 0.00012708689309657255, + "loss": 0.4322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42934340238571167, + "step": 3535, + "valid_targets_mean": 28909.0, + "valid_targets_min": 17573 + }, + { + "epoch": 3.7780149413020276, + "grad_norm": 0.4683631982078189, + "learning_rate": 0.00012686267607217954, + "loss": 0.4277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4246726632118225, + "step": 3540, + "valid_targets_mean": 28774.7, + "valid_targets_min": 17779 + }, + { + "epoch": 3.783351120597652, + "grad_norm": 0.3621086551412537, + "learning_rate": 0.00012663831340259222, + "loss": 0.4307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.425648033618927, + "step": 3545, + "valid_targets_mean": 28843.8, + "valid_targets_min": 22373 + }, + { + "epoch": 3.7886872998932764, + "grad_norm": 0.4451990706228817, + "learning_rate": 0.00012641380630426963, + "loss": 0.4307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4279008209705353, + "step": 3550, + "valid_targets_mean": 28949.8, + "valid_targets_min": 22022 + }, + { + "epoch": 3.7940234791889007, + "grad_norm": 0.4288991779222557, + "learning_rate": 0.00012618915599445387, + "loss": 0.4282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42941054701805115, + "step": 3555, + "valid_targets_mean": 28808.6, + "valid_targets_min": 17481 + }, + { + "epoch": 3.799359658484525, + "grad_norm": 0.39144371757232044, + "learning_rate": 0.00012596436369116348, + "loss": 0.4291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4281049966812134, + "step": 3560, + "valid_targets_mean": 28940.1, + "valid_targets_min": 22951 + }, + { + "epoch": 3.8046958377801494, + "grad_norm": 0.34309843042840693, + "learning_rate": 0.00012573943061318686, + "loss": 0.4276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4261920750141144, + "step": 3565, + "valid_targets_mean": 28768.5, + "valid_targets_min": 18305 + }, + { + "epoch": 3.8100320170757738, + "grad_norm": 0.42140784731783154, + "learning_rate": 0.00012551435798007572, + "loss": 0.4285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4318602383136749, + "step": 3570, + "valid_targets_mean": 28735.1, + "valid_targets_min": 21980 + }, + { + "epoch": 3.815368196371398, + "grad_norm": 0.5214249498817489, + "learning_rate": 0.0001252891470121384, + "loss": 0.4298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4317033886909485, + "step": 3575, + "valid_targets_mean": 29068.1, + "valid_targets_min": 22748 + }, + { + "epoch": 3.8207043756670225, + "grad_norm": 0.47742449376075385, + "learning_rate": 0.00012506379893043328, + "loss": 0.4302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43194687366485596, + "step": 3580, + "valid_targets_mean": 28783.8, + "valid_targets_min": 21027 + }, + { + "epoch": 3.826040554962647, + "grad_norm": 0.467160743712367, + "learning_rate": 0.00012483831495676206, + "loss": 0.4294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42947202920913696, + "step": 3585, + "valid_targets_mean": 28860.5, + "valid_targets_min": 23224 + }, + { + "epoch": 3.831376734258271, + "grad_norm": 0.4232740930262326, + "learning_rate": 0.00012461269631366338, + "loss": 0.4294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4264180064201355, + "step": 3590, + "valid_targets_mean": 28832.7, + "valid_targets_min": 16812 + }, + { + "epoch": 3.8367129135538955, + "grad_norm": 0.4080318700702358, + "learning_rate": 0.00012438694422440587, + "loss": 0.4293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.426061749458313, + "step": 3595, + "valid_targets_mean": 28853.5, + "valid_targets_min": 18470 + }, + { + "epoch": 3.84204909284952, + "grad_norm": 0.48170988188457003, + "learning_rate": 0.0001241610599129818, + "loss": 0.429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4270153045654297, + "step": 3600, + "valid_targets_mean": 28885.4, + "valid_targets_min": 21015 + }, + { + "epoch": 3.847385272145144, + "grad_norm": 0.3693315233172458, + "learning_rate": 0.0001239350446041003, + "loss": 0.4291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4311561584472656, + "step": 3605, + "valid_targets_mean": 28833.6, + "valid_targets_min": 23388 + }, + { + "epoch": 3.852721451440768, + "grad_norm": 0.3484926805840226, + "learning_rate": 0.00012370889952318076, + "loss": 0.4289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42880764603614807, + "step": 3610, + "valid_targets_mean": 28830.9, + "valid_targets_min": 21580 + }, + { + "epoch": 3.8580576307363925, + "grad_norm": 0.4335923435807277, + "learning_rate": 0.0001234826258963461, + "loss": 0.4289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42800524830818176, + "step": 3615, + "valid_targets_mean": 28794.4, + "valid_targets_min": 21666 + }, + { + "epoch": 3.863393810032017, + "grad_norm": 0.40222107492117587, + "learning_rate": 0.00012325622495041626, + "loss": 0.4304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4281304180622101, + "step": 3620, + "valid_targets_mean": 28763.7, + "valid_targets_min": 20819 + }, + { + "epoch": 3.8687299893276412, + "grad_norm": 0.3666339498052566, + "learning_rate": 0.00012302969791290152, + "loss": 0.4273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42554640769958496, + "step": 3625, + "valid_targets_mean": 28769.5, + "valid_targets_min": 17798 + }, + { + "epoch": 3.8740661686232656, + "grad_norm": 0.3378606634579012, + "learning_rate": 0.00012280304601199567, + "loss": 0.43, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4291246235370636, + "step": 3630, + "valid_targets_mean": 28854.0, + "valid_targets_min": 19509 + }, + { + "epoch": 3.87940234791889, + "grad_norm": 0.479120271244163, + "learning_rate": 0.0001225762704765697, + "loss": 0.4295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42979249358177185, + "step": 3635, + "valid_targets_mean": 28731.5, + "valid_targets_min": 21401 + }, + { + "epoch": 3.8847385272145143, + "grad_norm": 0.3862382523616369, + "learning_rate": 0.0001223493725361647, + "loss": 0.4283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42928624153137207, + "step": 3640, + "valid_targets_mean": 28895.1, + "valid_targets_min": 21913 + }, + { + "epoch": 3.8900747065101386, + "grad_norm": 0.42703226920414883, + "learning_rate": 0.00012212235342098563, + "loss": 0.429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42561978101730347, + "step": 3645, + "valid_targets_mean": 28699.3, + "valid_targets_min": 22699 + }, + { + "epoch": 3.895410885805763, + "grad_norm": 0.36739046228569433, + "learning_rate": 0.00012189521436189427, + "loss": 0.4274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4263993799686432, + "step": 3650, + "valid_targets_mean": 28909.2, + "valid_targets_min": 18879 + }, + { + "epoch": 3.9007470651013874, + "grad_norm": 0.3442698921561772, + "learning_rate": 0.00012166795659040278, + "loss": 0.429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4265468716621399, + "step": 3655, + "valid_targets_mean": 28808.9, + "valid_targets_min": 20693 + }, + { + "epoch": 3.9060832443970117, + "grad_norm": 0.3491683713142071, + "learning_rate": 0.00012144058133866696, + "loss": 0.4281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.427855908870697, + "step": 3660, + "valid_targets_mean": 28925.4, + "valid_targets_min": 22296 + }, + { + "epoch": 3.911419423692636, + "grad_norm": 0.39679933829431024, + "learning_rate": 0.00012121308983947963, + "loss": 0.4296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.430743932723999, + "step": 3665, + "valid_targets_mean": 28823.0, + "valid_targets_min": 19558 + }, + { + "epoch": 3.9167556029882604, + "grad_norm": 0.3808937270514705, + "learning_rate": 0.00012098548332626383, + "loss": 0.4278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42767736315727234, + "step": 3670, + "valid_targets_mean": 28807.6, + "valid_targets_min": 19538 + }, + { + "epoch": 3.9220917822838848, + "grad_norm": 0.4126907832167869, + "learning_rate": 0.00012075776303306612, + "loss": 0.428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42879045009613037, + "step": 3675, + "valid_targets_mean": 28829.6, + "valid_targets_min": 20457 + }, + { + "epoch": 3.927427961579509, + "grad_norm": 0.40542244391196997, + "learning_rate": 0.00012052993019455007, + "loss": 0.4252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4236496686935425, + "step": 3680, + "valid_targets_mean": 28944.2, + "valid_targets_min": 20050 + }, + { + "epoch": 3.9327641408751335, + "grad_norm": 0.43489706446937715, + "learning_rate": 0.00012030198604598943, + "loss": 0.4267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42238980531692505, + "step": 3685, + "valid_targets_mean": 28980.7, + "valid_targets_min": 21260 + }, + { + "epoch": 3.938100320170758, + "grad_norm": 0.4574175498837884, + "learning_rate": 0.00012007393182326136, + "loss": 0.4268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4301767945289612, + "step": 3690, + "valid_targets_mean": 28830.7, + "valid_targets_min": 20032 + }, + { + "epoch": 3.943436499466382, + "grad_norm": 0.5302303603287535, + "learning_rate": 0.00011984576876284, + "loss": 0.4276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4286668300628662, + "step": 3695, + "valid_targets_mean": 28710.5, + "valid_targets_min": 23020 + }, + { + "epoch": 3.9487726787620065, + "grad_norm": 0.42762805509272644, + "learning_rate": 0.00011961749810178943, + "loss": 0.4272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42783212661743164, + "step": 3700, + "valid_targets_mean": 28913.8, + "valid_targets_min": 21591 + }, + { + "epoch": 3.954108858057631, + "grad_norm": 0.3691856294166198, + "learning_rate": 0.00011938912107775717, + "loss": 0.4273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43037647008895874, + "step": 3705, + "valid_targets_mean": 29041.2, + "valid_targets_min": 23615 + }, + { + "epoch": 3.9594450373532553, + "grad_norm": 0.38810841873282675, + "learning_rate": 0.00011916063892896748, + "loss": 0.428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42605531215667725, + "step": 3710, + "valid_targets_mean": 28782.8, + "valid_targets_min": 19439 + }, + { + "epoch": 3.9647812166488796, + "grad_norm": 0.39822305684399667, + "learning_rate": 0.00011893205289421451, + "loss": 0.4277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42825180292129517, + "step": 3715, + "valid_targets_mean": 28692.7, + "valid_targets_min": 20633 + }, + { + "epoch": 3.970117395944504, + "grad_norm": 0.4158264081833239, + "learning_rate": 0.00011870336421285571, + "loss": 0.4286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4284394383430481, + "step": 3720, + "valid_targets_mean": 28782.1, + "valid_targets_min": 22529 + }, + { + "epoch": 3.9754535752401283, + "grad_norm": 0.3831811007794275, + "learning_rate": 0.00011847457412480505, + "loss": 0.426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42615097761154175, + "step": 3725, + "valid_targets_mean": 28755.5, + "valid_targets_min": 19295 + }, + { + "epoch": 3.9807897545357527, + "grad_norm": 0.4322480562579156, + "learning_rate": 0.00011824568387052637, + "loss": 0.4288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4323875904083252, + "step": 3730, + "valid_targets_mean": 28810.8, + "valid_targets_min": 19214 + }, + { + "epoch": 3.9861259338313766, + "grad_norm": 0.4694712272117525, + "learning_rate": 0.00011801669469102643, + "loss": 0.427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42549771070480347, + "step": 3735, + "valid_targets_mean": 28917.2, + "valid_targets_min": 22918 + }, + { + "epoch": 3.991462113127001, + "grad_norm": 0.3824415387016526, + "learning_rate": 0.00011778760782784855, + "loss": 0.4273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4290057122707367, + "step": 3740, + "valid_targets_mean": 28810.4, + "valid_targets_min": 19733 + }, + { + "epoch": 3.9967982924226253, + "grad_norm": 0.40822041699163486, + "learning_rate": 0.00011755842452306556, + "loss": 0.4273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42833641171455383, + "step": 3745, + "valid_targets_mean": 28969.8, + "valid_targets_min": 19636 + }, + { + "epoch": 4.00213447171825, + "grad_norm": 0.43314627908205433, + "learning_rate": 0.00011732914601927318, + "loss": 0.4239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4214213490486145, + "step": 3750, + "valid_targets_mean": 28785.9, + "valid_targets_min": 18581 + }, + { + "epoch": 4.0074706510138745, + "grad_norm": 0.4918842262705215, + "learning_rate": 0.00011709977355958333, + "loss": 0.4218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4230083227157593, + "step": 3755, + "valid_targets_mean": 29002.1, + "valid_targets_min": 22426 + }, + { + "epoch": 4.012806830309499, + "grad_norm": 0.3559040975229061, + "learning_rate": 0.00011687030838761732, + "loss": 0.4232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42185312509536743, + "step": 3760, + "valid_targets_mean": 28564.4, + "valid_targets_min": 18666 + }, + { + "epoch": 4.018143009605123, + "grad_norm": 0.3642085963533532, + "learning_rate": 0.00011664075174749911, + "loss": 0.4231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4223750829696655, + "step": 3765, + "valid_targets_mean": 28891.1, + "valid_targets_min": 23061 + }, + { + "epoch": 4.023479188900747, + "grad_norm": 0.3704274289489233, + "learning_rate": 0.00011641110488384866, + "loss": 0.4239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4247993528842926, + "step": 3770, + "valid_targets_mean": 28590.5, + "valid_targets_min": 19213 + }, + { + "epoch": 4.028815368196371, + "grad_norm": 0.3455884907079301, + "learning_rate": 0.00011618136904177501, + "loss": 0.4225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4216018319129944, + "step": 3775, + "valid_targets_mean": 28903.8, + "valid_targets_min": 22329 + }, + { + "epoch": 4.034151547491995, + "grad_norm": 0.3701656356814854, + "learning_rate": 0.00011595154546686969, + "loss": 0.4236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4252583086490631, + "step": 3780, + "valid_targets_mean": 28877.5, + "valid_targets_min": 22872 + }, + { + "epoch": 4.03948772678762, + "grad_norm": 0.3552312372354896, + "learning_rate": 0.00011572163540519986, + "loss": 0.4228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4176941514015198, + "step": 3785, + "valid_targets_mean": 28807.6, + "valid_targets_min": 20174 + }, + { + "epoch": 4.044823906083244, + "grad_norm": 0.41540332012123204, + "learning_rate": 0.00011549164010330171, + "loss": 0.4227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42301157116889954, + "step": 3790, + "valid_targets_mean": 28687.2, + "valid_targets_min": 20284 + }, + { + "epoch": 4.050160085378868, + "grad_norm": 0.4493140408504873, + "learning_rate": 0.00011526156080817341, + "loss": 0.4239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4221842885017395, + "step": 3795, + "valid_targets_mean": 28992.8, + "valid_targets_min": 21441 + }, + { + "epoch": 4.055496264674493, + "grad_norm": 0.38540924658447634, + "learning_rate": 0.00011503139876726863, + "loss": 0.4248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42516863346099854, + "step": 3800, + "valid_targets_mean": 28847.5, + "valid_targets_min": 20921 + }, + { + "epoch": 4.060832443970117, + "grad_norm": 0.3325527893251533, + "learning_rate": 0.00011480115522848973, + "loss": 0.4244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4216970205307007, + "step": 3805, + "valid_targets_mean": 28792.4, + "valid_targets_min": 19147 + }, + { + "epoch": 4.0661686232657415, + "grad_norm": 0.364134580511168, + "learning_rate": 0.00011457083144018083, + "loss": 0.422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4217662215232849, + "step": 3810, + "valid_targets_mean": 28953.0, + "valid_targets_min": 21609 + }, + { + "epoch": 4.071504802561366, + "grad_norm": 0.3685373613484052, + "learning_rate": 0.0001143404286511212, + "loss": 0.4231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42246201634407043, + "step": 3815, + "valid_targets_mean": 29024.4, + "valid_targets_min": 22957 + }, + { + "epoch": 4.07684098185699, + "grad_norm": 0.4020308844754011, + "learning_rate": 0.00011410994811051847, + "loss": 0.4214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42001721262931824, + "step": 3820, + "valid_targets_mean": 28833.5, + "valid_targets_min": 22585 + }, + { + "epoch": 4.0821771611526145, + "grad_norm": 0.42478665479006933, + "learning_rate": 0.00011387939106800176, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4186967611312866, + "step": 3825, + "valid_targets_mean": 28579.3, + "valid_targets_min": 20000 + }, + { + "epoch": 4.087513340448239, + "grad_norm": 0.39595288583207905, + "learning_rate": 0.00011364875877361496, + "loss": 0.4225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4193423390388489, + "step": 3830, + "valid_targets_mean": 28737.9, + "valid_targets_min": 22925 + }, + { + "epoch": 4.092849519743863, + "grad_norm": 0.4172708937499721, + "learning_rate": 0.00011341805247781008, + "loss": 0.4204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.424191415309906, + "step": 3835, + "valid_targets_mean": 28890.9, + "valid_targets_min": 23193 + }, + { + "epoch": 4.098185699039488, + "grad_norm": 0.3468813838294797, + "learning_rate": 0.00011318727343144026, + "loss": 0.4233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4230159819126129, + "step": 3840, + "valid_targets_mean": 28757.8, + "valid_targets_min": 22532 + }, + { + "epoch": 4.103521878335112, + "grad_norm": 0.3356951709404877, + "learning_rate": 0.00011295642288575307, + "loss": 0.4219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4207121729850769, + "step": 3845, + "valid_targets_mean": 28780.7, + "valid_targets_min": 19833 + }, + { + "epoch": 4.108858057630736, + "grad_norm": 0.39019958866424675, + "learning_rate": 0.00011272550209238387, + "loss": 0.4215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41962960362434387, + "step": 3850, + "valid_targets_mean": 28784.2, + "valid_targets_min": 22929 + }, + { + "epoch": 4.114194236926361, + "grad_norm": 0.4065746126214435, + "learning_rate": 0.00011249451230334867, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4226855933666229, + "step": 3855, + "valid_targets_mean": 28658.2, + "valid_targets_min": 21102 + }, + { + "epoch": 4.119530416221985, + "grad_norm": 0.3479054849722043, + "learning_rate": 0.00011226345477103773, + "loss": 0.4215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4219398498535156, + "step": 3860, + "valid_targets_mean": 28897.2, + "valid_targets_min": 23308 + }, + { + "epoch": 4.124866595517609, + "grad_norm": 0.348931221614668, + "learning_rate": 0.00011203233074820858, + "loss": 0.4216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41948723793029785, + "step": 3865, + "valid_targets_mean": 28637.0, + "valid_targets_min": 19468 + }, + { + "epoch": 4.130202774813234, + "grad_norm": 0.36593684546771144, + "learning_rate": 0.00011180114148797922, + "loss": 0.4237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42400026321411133, + "step": 3870, + "valid_targets_mean": 28747.5, + "valid_targets_min": 15351 + }, + { + "epoch": 4.135538954108858, + "grad_norm": 0.4094998797537509, + "learning_rate": 0.00011156988824382135, + "loss": 0.4228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4267050623893738, + "step": 3875, + "valid_targets_mean": 28884.7, + "valid_targets_min": 20049 + }, + { + "epoch": 4.140875133404482, + "grad_norm": 0.42888137632041157, + "learning_rate": 0.00011133857226955364, + "loss": 0.4209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42192575335502625, + "step": 3880, + "valid_targets_mean": 28725.5, + "valid_targets_min": 22448 + }, + { + "epoch": 4.146211312700107, + "grad_norm": 0.3826298745496494, + "learning_rate": 0.00011110719481933481, + "loss": 0.4235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4205487072467804, + "step": 3885, + "valid_targets_mean": 28963.6, + "valid_targets_min": 20484 + }, + { + "epoch": 4.151547491995731, + "grad_norm": 0.34768145283571444, + "learning_rate": 0.00011087575714765688, + "loss": 0.4234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42684638500213623, + "step": 3890, + "valid_targets_mean": 28793.0, + "valid_targets_min": 20882 + }, + { + "epoch": 4.1568836712913555, + "grad_norm": 0.4514800651060072, + "learning_rate": 0.00011064426050933847, + "loss": 0.4217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4261409044265747, + "step": 3895, + "valid_targets_mean": 28916.7, + "valid_targets_min": 20010 + }, + { + "epoch": 4.16221985058698, + "grad_norm": 0.33804355302083333, + "learning_rate": 0.00011041270615951782, + "loss": 0.4223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4239078164100647, + "step": 3900, + "valid_targets_mean": 28696.5, + "valid_targets_min": 21884 + }, + { + "epoch": 4.167556029882604, + "grad_norm": 0.4073212141419539, + "learning_rate": 0.00011018109535364613, + "loss": 0.421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42531710863113403, + "step": 3905, + "valid_targets_mean": 28859.9, + "valid_targets_min": 21850 + }, + { + "epoch": 4.172892209178229, + "grad_norm": 0.3976136524251578, + "learning_rate": 0.00010994942934748065, + "loss": 0.4211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42030707001686096, + "step": 3910, + "valid_targets_mean": 28898.3, + "valid_targets_min": 20825 + }, + { + "epoch": 4.178228388473853, + "grad_norm": 0.42144827688365827, + "learning_rate": 0.00010971770939707795, + "loss": 0.422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42350465059280396, + "step": 3915, + "valid_targets_mean": 28714.0, + "valid_targets_min": 20705 + }, + { + "epoch": 4.183564567769477, + "grad_norm": 0.3446658933506483, + "learning_rate": 0.00010948593675878706, + "loss": 0.4221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.424017995595932, + "step": 3920, + "valid_targets_mean": 28684.7, + "valid_targets_min": 21850 + }, + { + "epoch": 4.188900747065102, + "grad_norm": 0.4424627391052105, + "learning_rate": 0.00010925411268924268, + "loss": 0.4205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4225308299064636, + "step": 3925, + "valid_targets_mean": 28854.1, + "valid_targets_min": 22507 + }, + { + "epoch": 4.194236926360726, + "grad_norm": 0.37456413232780433, + "learning_rate": 0.00010902223844535839, + "loss": 0.4224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41586148738861084, + "step": 3930, + "valid_targets_mean": 28878.6, + "valid_targets_min": 21420 + }, + { + "epoch": 4.19957310565635, + "grad_norm": 0.3916860861831356, + "learning_rate": 0.00010879031528431975, + "loss": 0.42, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41759392619132996, + "step": 3935, + "valid_targets_mean": 28883.2, + "valid_targets_min": 22592 + }, + { + "epoch": 4.204909284951975, + "grad_norm": 0.37570024791802914, + "learning_rate": 0.0001085583444635776, + "loss": 0.423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.424510657787323, + "step": 3940, + "valid_targets_mean": 28893.8, + "valid_targets_min": 21433 + }, + { + "epoch": 4.210245464247599, + "grad_norm": 0.39543235368925594, + "learning_rate": 0.00010832632724084113, + "loss": 0.4217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41936933994293213, + "step": 3945, + "valid_targets_mean": 29016.8, + "valid_targets_min": 18748 + }, + { + "epoch": 4.215581643543223, + "grad_norm": 0.38859752670036407, + "learning_rate": 0.00010809426487407115, + "loss": 0.419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41779953241348267, + "step": 3950, + "valid_targets_mean": 28779.7, + "valid_targets_min": 15149 + }, + { + "epoch": 4.220917822838848, + "grad_norm": 0.3868922043328974, + "learning_rate": 0.0001078621586214732, + "loss": 0.4225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42176300287246704, + "step": 3955, + "valid_targets_mean": 28857.1, + "valid_targets_min": 21853 + }, + { + "epoch": 4.226254002134472, + "grad_norm": 0.31184443878340706, + "learning_rate": 0.00010763000974149084, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42075514793395996, + "step": 3960, + "valid_targets_mean": 28865.0, + "valid_targets_min": 20290 + }, + { + "epoch": 4.2315901814300965, + "grad_norm": 0.380274975601504, + "learning_rate": 0.00010739781949279865, + "loss": 0.4204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41909322142601013, + "step": 3965, + "valid_targets_mean": 28894.3, + "valid_targets_min": 21058 + }, + { + "epoch": 4.23692636072572, + "grad_norm": 0.4121484134744739, + "learning_rate": 0.00010716558913429557, + "loss": 0.4222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4200228452682495, + "step": 3970, + "valid_targets_mean": 28861.2, + "valid_targets_min": 21194 + }, + { + "epoch": 4.242262540021345, + "grad_norm": 0.4097411357038756, + "learning_rate": 0.00010693331992509801, + "loss": 0.4219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4192362427711487, + "step": 3975, + "valid_targets_mean": 28831.3, + "valid_targets_min": 21057 + }, + { + "epoch": 4.247598719316969, + "grad_norm": 0.4190985418435361, + "learning_rate": 0.00010670101312453293, + "loss": 0.4215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42083290219306946, + "step": 3980, + "valid_targets_mean": 28756.5, + "valid_targets_min": 21893 + }, + { + "epoch": 4.252934898612593, + "grad_norm": 0.4786577449254182, + "learning_rate": 0.00010646866999213123, + "loss": 0.4203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4252513647079468, + "step": 3985, + "valid_targets_mean": 28908.6, + "valid_targets_min": 20296 + }, + { + "epoch": 4.258271077908217, + "grad_norm": 0.3489029085337842, + "learning_rate": 0.00010623629178762073, + "loss": 0.4218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4199264645576477, + "step": 3990, + "valid_targets_mean": 28804.0, + "valid_targets_min": 20802 + }, + { + "epoch": 4.263607257203842, + "grad_norm": 0.4604933163875057, + "learning_rate": 0.00010600387977091944, + "loss": 0.4217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4209303557872772, + "step": 3995, + "valid_targets_mean": 28832.9, + "valid_targets_min": 22547 + }, + { + "epoch": 4.268943436499466, + "grad_norm": 0.38380160113267353, + "learning_rate": 0.00010577143520212865, + "loss": 0.4238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4269213080406189, + "step": 4000, + "valid_targets_mean": 28669.2, + "valid_targets_min": 19214 + }, + { + "epoch": 4.27427961579509, + "grad_norm": 0.40747812293986796, + "learning_rate": 0.00010553895934152613, + "loss": 0.4208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42184334993362427, + "step": 4005, + "valid_targets_mean": 28920.7, + "valid_targets_min": 17251 + }, + { + "epoch": 4.279615795090715, + "grad_norm": 0.3752844436503374, + "learning_rate": 0.0001053064534495594, + "loss": 0.4204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4206317067146301, + "step": 4010, + "valid_targets_mean": 28755.0, + "valid_targets_min": 22352 + }, + { + "epoch": 4.284951974386339, + "grad_norm": 0.4518199767432239, + "learning_rate": 0.00010507391878683867, + "loss": 0.4215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42242133617401123, + "step": 4015, + "valid_targets_mean": 28927.4, + "valid_targets_min": 22818 + }, + { + "epoch": 4.2902881536819635, + "grad_norm": 0.3475919231149287, + "learning_rate": 0.00010484135661413032, + "loss": 0.4209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4208674430847168, + "step": 4020, + "valid_targets_mean": 28732.6, + "valid_targets_min": 20510 + }, + { + "epoch": 4.295624332977588, + "grad_norm": 0.4235774556631322, + "learning_rate": 0.00010460876819234972, + "loss": 0.4221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4238291084766388, + "step": 4025, + "valid_targets_mean": 28766.6, + "valid_targets_min": 18717 + }, + { + "epoch": 4.300960512273212, + "grad_norm": 0.42439288693085914, + "learning_rate": 0.00010437615478255464, + "loss": 0.4199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41955137252807617, + "step": 4030, + "valid_targets_mean": 28905.3, + "valid_targets_min": 20991 + }, + { + "epoch": 4.3062966915688365, + "grad_norm": 0.4058575188077753, + "learning_rate": 0.0001041435176459383, + "loss": 0.4206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4235634207725525, + "step": 4035, + "valid_targets_mean": 29025.3, + "valid_targets_min": 22493 + }, + { + "epoch": 4.311632870864461, + "grad_norm": 0.347511414517583, + "learning_rate": 0.00010391085804382258, + "loss": 0.4193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4162161946296692, + "step": 4040, + "valid_targets_mean": 28853.1, + "valid_targets_min": 22919 + }, + { + "epoch": 4.316969050160085, + "grad_norm": 0.3575533250439514, + "learning_rate": 0.00010367817723765118, + "loss": 0.4178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41767430305480957, + "step": 4045, + "valid_targets_mean": 28817.6, + "valid_targets_min": 21231 + }, + { + "epoch": 4.32230522945571, + "grad_norm": 0.3776651010242567, + "learning_rate": 0.00010344547648898268, + "loss": 0.4212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41819173097610474, + "step": 4050, + "valid_targets_mean": 28915.1, + "valid_targets_min": 19795 + }, + { + "epoch": 4.327641408751334, + "grad_norm": 0.4060559950771988, + "learning_rate": 0.00010321275705948395, + "loss": 0.4208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.422052800655365, + "step": 4055, + "valid_targets_mean": 28808.5, + "valid_targets_min": 19111 + }, + { + "epoch": 4.332977588046958, + "grad_norm": 0.3279553395225615, + "learning_rate": 0.00010298002021092295, + "loss": 0.4211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42231857776641846, + "step": 4060, + "valid_targets_mean": 28935.0, + "valid_targets_min": 21699 + }, + { + "epoch": 4.338313767342583, + "grad_norm": 0.4027106897077968, + "learning_rate": 0.00010274726720516224, + "loss": 0.4222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4235973358154297, + "step": 4065, + "valid_targets_mean": 28856.1, + "valid_targets_min": 20560 + }, + { + "epoch": 4.343649946638207, + "grad_norm": 0.3173931093590618, + "learning_rate": 0.00010251449930415189, + "loss": 0.4211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42240917682647705, + "step": 4070, + "valid_targets_mean": 28815.0, + "valid_targets_min": 20116 + }, + { + "epoch": 4.348986125933831, + "grad_norm": 0.3570827061362892, + "learning_rate": 0.00010228171776992277, + "loss": 0.4198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42050933837890625, + "step": 4075, + "valid_targets_mean": 28649.3, + "valid_targets_min": 22583 + }, + { + "epoch": 4.354322305229456, + "grad_norm": 0.36354519154513204, + "learning_rate": 0.00010204892386457963, + "loss": 0.419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4194909632205963, + "step": 4080, + "valid_targets_mean": 28704.4, + "valid_targets_min": 20318 + }, + { + "epoch": 4.35965848452508, + "grad_norm": 0.39015491619158854, + "learning_rate": 0.00010181611885029437, + "loss": 0.4206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4181373417377472, + "step": 4085, + "valid_targets_mean": 28879.2, + "valid_targets_min": 19574 + }, + { + "epoch": 4.3649946638207044, + "grad_norm": 0.3839553114737321, + "learning_rate": 0.00010158330398929907, + "loss": 0.4217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4242491126060486, + "step": 4090, + "valid_targets_mean": 28708.3, + "valid_targets_min": 17719 + }, + { + "epoch": 4.370330843116329, + "grad_norm": 0.44199414398240855, + "learning_rate": 0.00010135048054387913, + "loss": 0.4208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42398786544799805, + "step": 4095, + "valid_targets_mean": 28836.2, + "valid_targets_min": 20619 + }, + { + "epoch": 4.375667022411953, + "grad_norm": 0.3618767207365597, + "learning_rate": 0.00010111764977636662, + "loss": 0.4198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42410895228385925, + "step": 4100, + "valid_targets_mean": 28895.4, + "valid_targets_min": 19004 + }, + { + "epoch": 4.3810032017075775, + "grad_norm": 0.36698779978922264, + "learning_rate": 0.00010088481294913325, + "loss": 0.4194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4190062880516052, + "step": 4105, + "valid_targets_mean": 28961.4, + "valid_targets_min": 22115 + }, + { + "epoch": 4.386339381003202, + "grad_norm": 0.3596786969412505, + "learning_rate": 0.00010065197132458355, + "loss": 0.4206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4156681001186371, + "step": 4110, + "valid_targets_mean": 29058.2, + "valid_targets_min": 23876 + }, + { + "epoch": 4.391675560298826, + "grad_norm": 0.3523444002878811, + "learning_rate": 0.00010041912616514814, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42014560103416443, + "step": 4115, + "valid_targets_mean": 28964.7, + "valid_targets_min": 21459 + }, + { + "epoch": 4.397011739594451, + "grad_norm": 0.40111593679590957, + "learning_rate": 0.00010018627873327677, + "loss": 0.4214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42068126797676086, + "step": 4120, + "valid_targets_mean": 28766.8, + "valid_targets_min": 20172 + }, + { + "epoch": 4.402347918890075, + "grad_norm": 0.40337010512586796, + "learning_rate": 9.995343029143144e-05, + "loss": 0.4204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41805556416511536, + "step": 4125, + "valid_targets_mean": 28770.8, + "valid_targets_min": 20890 + }, + { + "epoch": 4.407684098185699, + "grad_norm": 0.3658947575955679, + "learning_rate": 9.972058210207977e-05, + "loss": 0.42, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4142877757549286, + "step": 4130, + "valid_targets_mean": 28966.2, + "valid_targets_min": 20850 + }, + { + "epoch": 4.413020277481324, + "grad_norm": 0.30876775620978403, + "learning_rate": 9.948773542768784e-05, + "loss": 0.4184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41614896059036255, + "step": 4135, + "valid_targets_mean": 28884.3, + "valid_targets_min": 22054 + }, + { + "epoch": 4.418356456776948, + "grad_norm": 0.4322421389015756, + "learning_rate": 9.925489153071368e-05, + "loss": 0.4221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41993236541748047, + "step": 4140, + "valid_targets_mean": 28718.8, + "valid_targets_min": 21421 + }, + { + "epoch": 4.423692636072572, + "grad_norm": 0.3549190852580818, + "learning_rate": 9.902205167360015e-05, + "loss": 0.4185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4208112061023712, + "step": 4145, + "valid_targets_mean": 28854.2, + "valid_targets_min": 19793 + }, + { + "epoch": 4.429028815368197, + "grad_norm": 0.40680341585192115, + "learning_rate": 9.878921711876828e-05, + "loss": 0.4193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41954243183135986, + "step": 4150, + "valid_targets_mean": 28818.1, + "valid_targets_min": 18085 + }, + { + "epoch": 4.434364994663821, + "grad_norm": 0.32415051353565083, + "learning_rate": 9.855638912861026e-05, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4222108721733093, + "step": 4155, + "valid_targets_mean": 28776.8, + "valid_targets_min": 22162 + }, + { + "epoch": 4.439701173959445, + "grad_norm": 0.40712073556446066, + "learning_rate": 9.832356896548279e-05, + "loss": 0.4207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41958823800086975, + "step": 4160, + "valid_targets_mean": 28875.7, + "valid_targets_min": 19902 + }, + { + "epoch": 4.44503735325507, + "grad_norm": 0.326745826240599, + "learning_rate": 9.809075789170009e-05, + "loss": 0.4221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4246634840965271, + "step": 4165, + "valid_targets_mean": 28828.1, + "valid_targets_min": 17376 + }, + { + "epoch": 4.450373532550694, + "grad_norm": 0.3374293665013668, + "learning_rate": 9.785795716952704e-05, + "loss": 0.4205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4239436388015747, + "step": 4170, + "valid_targets_mean": 28768.5, + "valid_targets_min": 21819 + }, + { + "epoch": 4.455709711846318, + "grad_norm": 0.37830214672521734, + "learning_rate": 9.76251680611725e-05, + "loss": 0.4204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4206947684288025, + "step": 4175, + "valid_targets_mean": 28782.2, + "valid_targets_min": 21215 + }, + { + "epoch": 4.461045891141943, + "grad_norm": 0.3745343990155674, + "learning_rate": 9.739239182878228e-05, + "loss": 0.4195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4192778468132019, + "step": 4180, + "valid_targets_mean": 28579.8, + "valid_targets_min": 18932 + }, + { + "epoch": 4.466382070437566, + "grad_norm": 0.32899941991572457, + "learning_rate": 9.715962973443244e-05, + "loss": 0.4217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.423557311296463, + "step": 4185, + "valid_targets_mean": 28701.1, + "valid_targets_min": 20241 + }, + { + "epoch": 4.471718249733191, + "grad_norm": 0.2994773603134494, + "learning_rate": 9.692688304012239e-05, + "loss": 0.4199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4179522693157196, + "step": 4190, + "valid_targets_mean": 28791.0, + "valid_targets_min": 17497 + }, + { + "epoch": 4.477054429028815, + "grad_norm": 0.39070167349145646, + "learning_rate": 9.669415300776791e-05, + "loss": 0.4193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.421237975358963, + "step": 4195, + "valid_targets_mean": 28806.9, + "valid_targets_min": 17517 + }, + { + "epoch": 4.482390608324439, + "grad_norm": 0.3398067579522288, + "learning_rate": 9.646144089919462e-05, + "loss": 0.4211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4213641285896301, + "step": 4200, + "valid_targets_mean": 28845.2, + "valid_targets_min": 23682 + }, + { + "epoch": 4.487726787620064, + "grad_norm": 0.36298214897813275, + "learning_rate": 9.622874797613084e-05, + "loss": 0.4186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4160724878311157, + "step": 4205, + "valid_targets_mean": 28931.6, + "valid_targets_min": 20098 + }, + { + "epoch": 4.493062966915688, + "grad_norm": 0.3358526558198807, + "learning_rate": 9.599607550020097e-05, + "loss": 0.4201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41969814896583557, + "step": 4210, + "valid_targets_mean": 28897.4, + "valid_targets_min": 23228 + }, + { + "epoch": 4.498399146211312, + "grad_norm": 0.292914484584559, + "learning_rate": 9.576342473291842e-05, + "loss": 0.4208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4190637469291687, + "step": 4215, + "valid_targets_mean": 28789.5, + "valid_targets_min": 22550 + }, + { + "epoch": 4.503735325506937, + "grad_norm": 0.37867376396765706, + "learning_rate": 9.5530796935679e-05, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4196445345878601, + "step": 4220, + "valid_targets_mean": 28898.7, + "valid_targets_min": 21128 + }, + { + "epoch": 4.509071504802561, + "grad_norm": 0.34320896429190056, + "learning_rate": 9.529819336975399e-05, + "loss": 0.4192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4201467037200928, + "step": 4225, + "valid_targets_mean": 28947.5, + "valid_targets_min": 21968 + }, + { + "epoch": 4.5144076840981855, + "grad_norm": 0.3345778358392203, + "learning_rate": 9.506561529628315e-05, + "loss": 0.4196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4164847135543823, + "step": 4230, + "valid_targets_mean": 28720.5, + "valid_targets_min": 19838 + }, + { + "epoch": 4.51974386339381, + "grad_norm": 0.42696060130478747, + "learning_rate": 9.48330639762682e-05, + "loss": 0.4201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4239192605018616, + "step": 4235, + "valid_targets_mean": 28780.6, + "valid_targets_min": 21973 + }, + { + "epoch": 4.525080042689434, + "grad_norm": 0.3493891285321995, + "learning_rate": 9.46005406705657e-05, + "loss": 0.4178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41531574726104736, + "step": 4240, + "valid_targets_mean": 28854.0, + "valid_targets_min": 20820 + }, + { + "epoch": 4.5304162219850586, + "grad_norm": 0.4017298856118809, + "learning_rate": 9.436804663988041e-05, + "loss": 0.4208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4187561273574829, + "step": 4245, + "valid_targets_mean": 28727.1, + "valid_targets_min": 20099 + }, + { + "epoch": 4.535752401280683, + "grad_norm": 0.34885821775638554, + "learning_rate": 9.413558314475825e-05, + "loss": 0.4213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4193626046180725, + "step": 4250, + "valid_targets_mean": 28871.8, + "valid_targets_min": 19352 + }, + { + "epoch": 4.541088580576307, + "grad_norm": 0.3550438111233903, + "learning_rate": 9.390315144557961e-05, + "loss": 0.4207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41961002349853516, + "step": 4255, + "valid_targets_mean": 28847.8, + "valid_targets_min": 21715 + }, + { + "epoch": 4.546424759871932, + "grad_norm": 0.35789280524944217, + "learning_rate": 9.367075280255262e-05, + "loss": 0.4187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41878896951675415, + "step": 4260, + "valid_targets_mean": 29012.0, + "valid_targets_min": 20482 + }, + { + "epoch": 4.551760939167556, + "grad_norm": 0.3234103601058616, + "learning_rate": 9.343838847570594e-05, + "loss": 0.4202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42339909076690674, + "step": 4265, + "valid_targets_mean": 28761.8, + "valid_targets_min": 21638 + }, + { + "epoch": 4.55709711846318, + "grad_norm": 0.392164466338175, + "learning_rate": 9.320605972488247e-05, + "loss": 0.4192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41794681549072266, + "step": 4270, + "valid_targets_mean": 28815.6, + "valid_targets_min": 18518 + }, + { + "epoch": 4.562433297758805, + "grad_norm": 0.3262995318235304, + "learning_rate": 9.297376780973192e-05, + "loss": 0.4186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41663309931755066, + "step": 4275, + "valid_targets_mean": 28847.7, + "valid_targets_min": 19122 + }, + { + "epoch": 4.567769477054429, + "grad_norm": 0.316402610712325, + "learning_rate": 9.274151398970456e-05, + "loss": 0.4191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4165911078453064, + "step": 4280, + "valid_targets_mean": 28901.4, + "valid_targets_min": 22373 + }, + { + "epoch": 4.573105656350053, + "grad_norm": 0.36717371358448186, + "learning_rate": 9.250929952404392e-05, + "loss": 0.4166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41654711961746216, + "step": 4285, + "valid_targets_mean": 28828.3, + "valid_targets_min": 18039 + }, + { + "epoch": 4.578441835645678, + "grad_norm": 0.3160343590979992, + "learning_rate": 9.227712567178018e-05, + "loss": 0.4204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42192721366882324, + "step": 4290, + "valid_targets_mean": 28948.7, + "valid_targets_min": 22103 + }, + { + "epoch": 4.583778014941302, + "grad_norm": 0.3516830197853869, + "learning_rate": 9.204499369172343e-05, + "loss": 0.4198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.422343373298645, + "step": 4295, + "valid_targets_mean": 28722.5, + "valid_targets_min": 20948 + }, + { + "epoch": 4.5891141942369265, + "grad_norm": 0.37186306459070556, + "learning_rate": 9.18129048424566e-05, + "loss": 0.4197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42082688212394714, + "step": 4300, + "valid_targets_mean": 29034.9, + "valid_targets_min": 18977 + }, + { + "epoch": 4.594450373532551, + "grad_norm": 0.3229935606530557, + "learning_rate": 9.158086038232894e-05, + "loss": 0.4194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4219149649143219, + "step": 4305, + "valid_targets_mean": 28729.0, + "valid_targets_min": 19451 + }, + { + "epoch": 4.599786552828175, + "grad_norm": 0.3579114181012708, + "learning_rate": 9.134886156944883e-05, + "loss": 0.4205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.424641489982605, + "step": 4310, + "valid_targets_mean": 28805.3, + "valid_targets_min": 17376 + }, + { + "epoch": 4.6051227321237995, + "grad_norm": 0.4211346528799389, + "learning_rate": 9.111690966167727e-05, + "loss": 0.4205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42049410939216614, + "step": 4315, + "valid_targets_mean": 28814.6, + "valid_targets_min": 17756 + }, + { + "epoch": 4.610458911419424, + "grad_norm": 0.3418778948184453, + "learning_rate": 9.088500591662099e-05, + "loss": 0.4191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41268154978752136, + "step": 4320, + "valid_targets_mean": 28675.5, + "valid_targets_min": 16701 + }, + { + "epoch": 4.615795090715048, + "grad_norm": 0.3409962577347365, + "learning_rate": 9.065315159162546e-05, + "loss": 0.4203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4228631556034088, + "step": 4325, + "valid_targets_mean": 28863.2, + "valid_targets_min": 19723 + }, + { + "epoch": 4.621131270010673, + "grad_norm": 0.2805196210087388, + "learning_rate": 9.042134794376833e-05, + "loss": 0.4196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41876015067100525, + "step": 4330, + "valid_targets_mean": 28818.4, + "valid_targets_min": 21421 + }, + { + "epoch": 4.626467449306297, + "grad_norm": 0.36554232885208404, + "learning_rate": 9.018959622985238e-05, + "loss": 0.4166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4189041554927826, + "step": 4335, + "valid_targets_mean": 28728.7, + "valid_targets_min": 20241 + }, + { + "epoch": 4.631803628601921, + "grad_norm": 0.325278718856548, + "learning_rate": 8.99578977063989e-05, + "loss": 0.4178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.413724422454834, + "step": 4340, + "valid_targets_mean": 28706.4, + "valid_targets_min": 18963 + }, + { + "epoch": 4.637139807897546, + "grad_norm": 0.3431558034381059, + "learning_rate": 8.972625362964078e-05, + "loss": 0.4186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4200518727302551, + "step": 4345, + "valid_targets_mean": 28807.0, + "valid_targets_min": 18152 + }, + { + "epoch": 4.64247598719317, + "grad_norm": 0.3936183086017574, + "learning_rate": 8.949466525551561e-05, + "loss": 0.4196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4186498522758484, + "step": 4350, + "valid_targets_mean": 28906.8, + "valid_targets_min": 22116 + }, + { + "epoch": 4.647812166488794, + "grad_norm": 0.2589150848327681, + "learning_rate": 8.92631338396591e-05, + "loss": 0.4194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41684263944625854, + "step": 4355, + "valid_targets_mean": 28874.1, + "valid_targets_min": 21201 + }, + { + "epoch": 4.653148345784419, + "grad_norm": 0.42417381425709433, + "learning_rate": 8.903166063739801e-05, + "loss": 0.4196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4187427759170532, + "step": 4360, + "valid_targets_mean": 28964.5, + "valid_targets_min": 21909 + }, + { + "epoch": 4.658484525080043, + "grad_norm": 0.418946503334357, + "learning_rate": 8.880024690374367e-05, + "loss": 0.4206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4185762405395508, + "step": 4365, + "valid_targets_mean": 28706.8, + "valid_targets_min": 20141 + }, + { + "epoch": 4.6638207043756665, + "grad_norm": 0.3436223121999799, + "learning_rate": 8.856889389338474e-05, + "loss": 0.4173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41747280955314636, + "step": 4370, + "valid_targets_mean": 28651.3, + "valid_targets_min": 19727 + }, + { + "epoch": 4.669156883671292, + "grad_norm": 0.3089754644534319, + "learning_rate": 8.833760286068079e-05, + "loss": 0.4184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4154074192047119, + "step": 4375, + "valid_targets_mean": 28847.3, + "valid_targets_min": 21510 + }, + { + "epoch": 4.674493062966915, + "grad_norm": 0.35073468955471265, + "learning_rate": 8.81063750596554e-05, + "loss": 0.4188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4182649254798889, + "step": 4380, + "valid_targets_mean": 28780.5, + "valid_targets_min": 22501 + }, + { + "epoch": 4.6798292422625405, + "grad_norm": 0.3380932879099673, + "learning_rate": 8.787521174398914e-05, + "loss": 0.4175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4169265329837799, + "step": 4385, + "valid_targets_mean": 28870.8, + "valid_targets_min": 22732 + }, + { + "epoch": 4.685165421558164, + "grad_norm": 0.33067693590054464, + "learning_rate": 8.764411416701317e-05, + "loss": 0.4186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4172070622444153, + "step": 4390, + "valid_targets_mean": 28932.9, + "valid_targets_min": 22018 + }, + { + "epoch": 4.690501600853788, + "grad_norm": 0.27120664884307055, + "learning_rate": 8.741308358170202e-05, + "loss": 0.4201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42023521661758423, + "step": 4395, + "valid_targets_mean": 29050.9, + "valid_targets_min": 22707 + }, + { + "epoch": 4.695837780149413, + "grad_norm": 0.3965087073853456, + "learning_rate": 8.71821212406672e-05, + "loss": 0.4184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4160880446434021, + "step": 4400, + "valid_targets_mean": 28808.7, + "valid_targets_min": 21127 + }, + { + "epoch": 4.701173959445037, + "grad_norm": 0.29841783707385844, + "learning_rate": 8.695122839615004e-05, + "loss": 0.4169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41510114073753357, + "step": 4405, + "valid_targets_mean": 28624.5, + "valid_targets_min": 20551 + }, + { + "epoch": 4.706510138740661, + "grad_norm": 0.33245364871492056, + "learning_rate": 8.672040630001511e-05, + "loss": 0.4178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41986995935440063, + "step": 4410, + "valid_targets_mean": 28915.2, + "valid_targets_min": 22626 + }, + { + "epoch": 4.711846318036286, + "grad_norm": 0.306725569562423, + "learning_rate": 8.648965620374352e-05, + "loss": 0.4188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4156710207462311, + "step": 4415, + "valid_targets_mean": 28743.1, + "valid_targets_min": 13210 + }, + { + "epoch": 4.71718249733191, + "grad_norm": 0.37738100167382255, + "learning_rate": 8.625897935842578e-05, + "loss": 0.4176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4194144010543823, + "step": 4420, + "valid_targets_mean": 28895.6, + "valid_targets_min": 22626 + }, + { + "epoch": 4.722518676627534, + "grad_norm": 0.34450445819147957, + "learning_rate": 8.602837701475551e-05, + "loss": 0.4171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.412597119808197, + "step": 4425, + "valid_targets_mean": 28783.0, + "valid_targets_min": 22626 + }, + { + "epoch": 4.727854855923159, + "grad_norm": 0.3038540518074493, + "learning_rate": 8.579785042302218e-05, + "loss": 0.4168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4147343635559082, + "step": 4430, + "valid_targets_mean": 28978.0, + "valid_targets_min": 22913 + }, + { + "epoch": 4.733191035218783, + "grad_norm": 0.44131096389065855, + "learning_rate": 8.556740083310463e-05, + "loss": 0.4191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4168371558189392, + "step": 4435, + "valid_targets_mean": 28870.3, + "valid_targets_min": 21449 + }, + { + "epoch": 4.7385272145144075, + "grad_norm": 0.39415552699192785, + "learning_rate": 8.533702949446424e-05, + "loss": 0.4176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41771623492240906, + "step": 4440, + "valid_targets_mean": 28764.5, + "valid_targets_min": 21786 + }, + { + "epoch": 4.743863393810032, + "grad_norm": 0.3287126670187571, + "learning_rate": 8.510673765613803e-05, + "loss": 0.4206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4177106022834778, + "step": 4445, + "valid_targets_mean": 28957.6, + "valid_targets_min": 20693 + }, + { + "epoch": 4.749199573105656, + "grad_norm": 0.28632108037662457, + "learning_rate": 8.487652656673209e-05, + "loss": 0.4158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4170299172401428, + "step": 4450, + "valid_targets_mean": 28610.0, + "valid_targets_min": 21408 + }, + { + "epoch": 4.754535752401281, + "grad_norm": 0.3122943893226229, + "learning_rate": 8.464639747441459e-05, + "loss": 0.4186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41472533345222473, + "step": 4455, + "valid_targets_mean": 28899.9, + "valid_targets_min": 20112 + }, + { + "epoch": 4.759871931696905, + "grad_norm": 0.2767625756846586, + "learning_rate": 8.441635162690927e-05, + "loss": 0.4178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4218922257423401, + "step": 4460, + "valid_targets_mean": 29026.5, + "valid_targets_min": 23953 + }, + { + "epoch": 4.765208110992529, + "grad_norm": 0.3016890644467839, + "learning_rate": 8.418639027148838e-05, + "loss": 0.418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41578352451324463, + "step": 4465, + "valid_targets_mean": 28831.1, + "valid_targets_min": 20621 + }, + { + "epoch": 4.770544290288154, + "grad_norm": 0.32405861806307124, + "learning_rate": 8.395651465496613e-05, + "loss": 0.4169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41797369718551636, + "step": 4470, + "valid_targets_mean": 28837.1, + "valid_targets_min": 19761 + }, + { + "epoch": 4.775880469583778, + "grad_norm": 0.30666236660222157, + "learning_rate": 8.372672602369191e-05, + "loss": 0.4153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41425567865371704, + "step": 4475, + "valid_targets_mean": 28974.5, + "valid_targets_min": 21885 + }, + { + "epoch": 4.781216648879402, + "grad_norm": 0.2961485972394307, + "learning_rate": 8.34970256235434e-05, + "loss": 0.4179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4199238717556, + "step": 4480, + "valid_targets_mean": 28834.1, + "valid_targets_min": 21855 + }, + { + "epoch": 4.786552828175027, + "grad_norm": 0.29191812028156455, + "learning_rate": 8.326741469992001e-05, + "loss": 0.4165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4175301194190979, + "step": 4485, + "valid_targets_mean": 28841.4, + "valid_targets_min": 21643 + }, + { + "epoch": 4.791889007470651, + "grad_norm": 0.3043402658104164, + "learning_rate": 8.303789449773593e-05, + "loss": 0.4175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.418673038482666, + "step": 4490, + "valid_targets_mean": 28827.9, + "valid_targets_min": 21275 + }, + { + "epoch": 4.797225186766275, + "grad_norm": 0.29684399099924325, + "learning_rate": 8.28084662614135e-05, + "loss": 0.4182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4179407060146332, + "step": 4495, + "valid_targets_mean": 28900.9, + "valid_targets_min": 22257 + }, + { + "epoch": 4.8025613660619, + "grad_norm": 0.3068560334729122, + "learning_rate": 8.257913123487646e-05, + "loss": 0.4169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41665366291999817, + "step": 4500, + "valid_targets_mean": 28967.9, + "valid_targets_min": 22325 + }, + { + "epoch": 4.807897545357524, + "grad_norm": 0.383704390382649, + "learning_rate": 8.234989066154314e-05, + "loss": 0.4184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41750168800354004, + "step": 4505, + "valid_targets_mean": 28859.4, + "valid_targets_min": 23095 + }, + { + "epoch": 4.8132337246531485, + "grad_norm": 0.31464696881108956, + "learning_rate": 8.212074578431983e-05, + "loss": 0.4152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41778695583343506, + "step": 4510, + "valid_targets_mean": 28916.0, + "valid_targets_min": 18485 + }, + { + "epoch": 4.818569903948773, + "grad_norm": 0.3401817851623499, + "learning_rate": 8.189169784559388e-05, + "loss": 0.4176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4168986678123474, + "step": 4515, + "valid_targets_mean": 28711.9, + "valid_targets_min": 21349 + }, + { + "epoch": 4.823906083244397, + "grad_norm": 0.31446079438830254, + "learning_rate": 8.166274808722717e-05, + "loss": 0.416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41559553146362305, + "step": 4520, + "valid_targets_mean": 28771.6, + "valid_targets_min": 20055 + }, + { + "epoch": 4.8292422625400215, + "grad_norm": 0.37148698409607017, + "learning_rate": 8.143389775054913e-05, + "loss": 0.4144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4131711721420288, + "step": 4525, + "valid_targets_mean": 28967.4, + "valid_targets_min": 23007 + }, + { + "epoch": 4.834578441835646, + "grad_norm": 0.40259768703683657, + "learning_rate": 8.120514807635019e-05, + "loss": 0.4166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41874104738235474, + "step": 4530, + "valid_targets_mean": 28941.8, + "valid_targets_min": 21343 + }, + { + "epoch": 4.83991462113127, + "grad_norm": 0.3111587866730002, + "learning_rate": 8.097650030487509e-05, + "loss": 0.4174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41815048456192017, + "step": 4535, + "valid_targets_mean": 28776.4, + "valid_targets_min": 17463 + }, + { + "epoch": 4.845250800426895, + "grad_norm": 0.4037662210760224, + "learning_rate": 8.07479556758159e-05, + "loss": 0.4187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4164711833000183, + "step": 4540, + "valid_targets_mean": 28861.0, + "valid_targets_min": 21525 + }, + { + "epoch": 4.850586979722519, + "grad_norm": 0.3653771293856302, + "learning_rate": 8.051951542830565e-05, + "loss": 0.4174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4157501459121704, + "step": 4545, + "valid_targets_mean": 28876.5, + "valid_targets_min": 22176 + }, + { + "epoch": 4.855923159018143, + "grad_norm": 0.24448778166438007, + "learning_rate": 8.029118080091133e-05, + "loss": 0.417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4192211627960205, + "step": 4550, + "valid_targets_mean": 28988.9, + "valid_targets_min": 19752 + }, + { + "epoch": 4.861259338313768, + "grad_norm": 0.3065192095331345, + "learning_rate": 8.006295303162723e-05, + "loss": 0.4168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4171658456325531, + "step": 4555, + "valid_targets_mean": 28943.8, + "valid_targets_min": 18260 + }, + { + "epoch": 4.866595517609392, + "grad_norm": 0.34152052598085214, + "learning_rate": 7.983483335786838e-05, + "loss": 0.4174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41691333055496216, + "step": 4560, + "valid_targets_mean": 28818.9, + "valid_targets_min": 19534 + }, + { + "epoch": 4.871931696905016, + "grad_norm": 0.34423404010059494, + "learning_rate": 7.960682301646367e-05, + "loss": 0.4174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4178677499294281, + "step": 4565, + "valid_targets_mean": 28691.4, + "valid_targets_min": 21732 + }, + { + "epoch": 4.877267876200641, + "grad_norm": 0.3502990969830706, + "learning_rate": 7.937892324364927e-05, + "loss": 0.4167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41543686389923096, + "step": 4570, + "valid_targets_mean": 28915.1, + "valid_targets_min": 21491 + }, + { + "epoch": 4.882604055496264, + "grad_norm": 0.3067683857912805, + "learning_rate": 7.915113527506172e-05, + "loss": 0.4191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41942209005355835, + "step": 4575, + "valid_targets_mean": 29039.0, + "valid_targets_min": 23137 + }, + { + "epoch": 4.887940234791889, + "grad_norm": 0.2858703834942536, + "learning_rate": 7.892346034573155e-05, + "loss": 0.4146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41157788038253784, + "step": 4580, + "valid_targets_mean": 28865.8, + "valid_targets_min": 21471 + }, + { + "epoch": 4.893276414087513, + "grad_norm": 0.3073279216029923, + "learning_rate": 7.869589969007631e-05, + "loss": 0.419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.418150931596756, + "step": 4585, + "valid_targets_mean": 28745.0, + "valid_targets_min": 22147 + }, + { + "epoch": 4.898612593383138, + "grad_norm": 0.2578006093415181, + "learning_rate": 7.846845454189395e-05, + "loss": 0.4169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41997313499450684, + "step": 4590, + "valid_targets_mean": 28904.9, + "valid_targets_min": 23113 + }, + { + "epoch": 4.903948772678762, + "grad_norm": 0.31226067833635385, + "learning_rate": 7.82411261343563e-05, + "loss": 0.4167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41901105642318726, + "step": 4595, + "valid_targets_mean": 29030.1, + "valid_targets_min": 22724 + }, + { + "epoch": 4.909284951974386, + "grad_norm": 0.30264006623904355, + "learning_rate": 7.801391570000203e-05, + "loss": 0.4156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4186111092567444, + "step": 4600, + "valid_targets_mean": 28866.0, + "valid_targets_min": 19994 + }, + { + "epoch": 4.91462113127001, + "grad_norm": 0.30406421360587316, + "learning_rate": 7.778682447073037e-05, + "loss": 0.4165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4153941869735718, + "step": 4605, + "valid_targets_mean": 28828.0, + "valid_targets_min": 21703 + }, + { + "epoch": 4.919957310565635, + "grad_norm": 0.2833979268310583, + "learning_rate": 7.755985367779417e-05, + "loss": 0.416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41638970375061035, + "step": 4610, + "valid_targets_mean": 28857.7, + "valid_targets_min": 18061 + }, + { + "epoch": 4.925293489861259, + "grad_norm": 0.3102148343129894, + "learning_rate": 7.73330045517932e-05, + "loss": 0.418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4176219403743744, + "step": 4615, + "valid_targets_mean": 28663.0, + "valid_targets_min": 20634 + }, + { + "epoch": 4.930629669156883, + "grad_norm": 0.2785617541092793, + "learning_rate": 7.710627832266772e-05, + "loss": 0.4181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4157201945781708, + "step": 4620, + "valid_targets_mean": 28889.3, + "valid_targets_min": 17762 + }, + { + "epoch": 4.935965848452508, + "grad_norm": 0.3049573179938812, + "learning_rate": 7.687967621969156e-05, + "loss": 0.4154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4141373336315155, + "step": 4625, + "valid_targets_mean": 28693.4, + "valid_targets_min": 19540 + }, + { + "epoch": 4.941302027748132, + "grad_norm": 0.2938453993328095, + "learning_rate": 7.665319947146567e-05, + "loss": 0.419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4159335494041443, + "step": 4630, + "valid_targets_mean": 28816.9, + "valid_targets_min": 21054 + }, + { + "epoch": 4.9466382070437565, + "grad_norm": 0.33632792903028347, + "learning_rate": 7.642684930591118e-05, + "loss": 0.4147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4127083122730255, + "step": 4635, + "valid_targets_mean": 28978.0, + "valid_targets_min": 22893 + }, + { + "epoch": 4.951974386339381, + "grad_norm": 0.3439403913153363, + "learning_rate": 7.620062695026305e-05, + "loss": 0.4144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4118298590183258, + "step": 4640, + "valid_targets_mean": 28901.1, + "valid_targets_min": 22439 + }, + { + "epoch": 4.957310565635005, + "grad_norm": 0.29311156891677753, + "learning_rate": 7.597453363106323e-05, + "loss": 0.4177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4171184301376343, + "step": 4645, + "valid_targets_mean": 28937.9, + "valid_targets_min": 19838 + }, + { + "epoch": 4.9626467449306295, + "grad_norm": 0.2908629315122143, + "learning_rate": 7.574857057415401e-05, + "loss": 0.4144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4137275218963623, + "step": 4650, + "valid_targets_mean": 28969.2, + "valid_targets_min": 21641 + }, + { + "epoch": 4.967982924226254, + "grad_norm": 0.30981310592678857, + "learning_rate": 7.552273900467149e-05, + "loss": 0.4154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4096093475818634, + "step": 4655, + "valid_targets_mean": 28731.8, + "valid_targets_min": 22058 + }, + { + "epoch": 4.973319103521878, + "grad_norm": 0.30163436028418267, + "learning_rate": 7.529704014703878e-05, + "loss": 0.4175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41557684540748596, + "step": 4660, + "valid_targets_mean": 28835.5, + "valid_targets_min": 20865 + }, + { + "epoch": 4.978655282817503, + "grad_norm": 0.32491627154320946, + "learning_rate": 7.50714752249596e-05, + "loss": 0.4155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41583991050720215, + "step": 4665, + "valid_targets_mean": 28765.9, + "valid_targets_min": 21199 + }, + { + "epoch": 4.983991462113127, + "grad_norm": 0.2716746089249112, + "learning_rate": 7.484604546141132e-05, + "loss": 0.4163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4182189404964447, + "step": 4670, + "valid_targets_mean": 28768.7, + "valid_targets_min": 18559 + }, + { + "epoch": 4.989327641408751, + "grad_norm": 0.317168435680314, + "learning_rate": 7.462075207863856e-05, + "loss": 0.4154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41550585627555847, + "step": 4675, + "valid_targets_mean": 28763.5, + "valid_targets_min": 23064 + }, + { + "epoch": 4.994663820704376, + "grad_norm": 0.2869849007273826, + "learning_rate": 7.439559629814657e-05, + "loss": 0.4143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4126055836677551, + "step": 4680, + "valid_targets_mean": 28877.9, + "valid_targets_min": 22243 + }, + { + "epoch": 5.0, + "grad_norm": 0.3184488694647241, + "learning_rate": 7.417057934069445e-05, + "loss": 0.4132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41238728165626526, + "step": 4685, + "valid_targets_mean": 28912.7, + "valid_targets_min": 17388 + }, + { + "epoch": 5.005336179295624, + "grad_norm": 0.2925300189389251, + "learning_rate": 7.394570242628875e-05, + "loss": 0.4111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41013145446777344, + "step": 4690, + "valid_targets_mean": 28927.0, + "valid_targets_min": 20840 + }, + { + "epoch": 5.010672358591249, + "grad_norm": 0.32574966258870014, + "learning_rate": 7.372096677417652e-05, + "loss": 0.4106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41071006655693054, + "step": 4695, + "valid_targets_mean": 28954.2, + "valid_targets_min": 22117 + }, + { + "epoch": 5.016008537886873, + "grad_norm": 0.3246810617001424, + "learning_rate": 7.349637360283912e-05, + "loss": 0.4112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.410560667514801, + "step": 4700, + "valid_targets_mean": 28753.2, + "valid_targets_min": 22469 + }, + { + "epoch": 5.021344717182497, + "grad_norm": 0.32843109233372075, + "learning_rate": 7.327192412998531e-05, + "loss": 0.4119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4146524667739868, + "step": 4705, + "valid_targets_mean": 28796.6, + "valid_targets_min": 23038 + }, + { + "epoch": 5.026680896478122, + "grad_norm": 0.32243355206822194, + "learning_rate": 7.304761957254468e-05, + "loss": 0.4126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41325482726097107, + "step": 4710, + "valid_targets_mean": 28702.6, + "valid_targets_min": 18678 + }, + { + "epoch": 5.032017075773746, + "grad_norm": 0.30617066599895376, + "learning_rate": 7.282346114666125e-05, + "loss": 0.4116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41367843747138977, + "step": 4715, + "valid_targets_mean": 28808.5, + "valid_targets_min": 21490 + }, + { + "epoch": 5.0373532550693705, + "grad_norm": 0.2622483401127021, + "learning_rate": 7.259945006768661e-05, + "loss": 0.4102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40470439195632935, + "step": 4720, + "valid_targets_mean": 28830.8, + "valid_targets_min": 20938 + }, + { + "epoch": 5.042689434364995, + "grad_norm": 0.24084554811633835, + "learning_rate": 7.237558755017358e-05, + "loss": 0.4132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41252872347831726, + "step": 4725, + "valid_targets_mean": 28893.2, + "valid_targets_min": 21118 + }, + { + "epoch": 5.048025613660619, + "grad_norm": 0.3412153342192512, + "learning_rate": 7.215187480786934e-05, + "loss": 0.4111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41102367639541626, + "step": 4730, + "valid_targets_mean": 28947.7, + "valid_targets_min": 19871 + }, + { + "epoch": 5.0533617929562435, + "grad_norm": 0.25635523803675986, + "learning_rate": 7.192831305370913e-05, + "loss": 0.4114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4131872057914734, + "step": 4735, + "valid_targets_mean": 28995.2, + "valid_targets_min": 20824 + }, + { + "epoch": 5.058697972251868, + "grad_norm": 0.3469477127452649, + "learning_rate": 7.170490349980961e-05, + "loss": 0.4105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4078807830810547, + "step": 4740, + "valid_targets_mean": 28730.4, + "valid_targets_min": 19013 + }, + { + "epoch": 5.064034151547492, + "grad_norm": 0.31461141347673727, + "learning_rate": 7.148164735746206e-05, + "loss": 0.4103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41446444392204285, + "step": 4745, + "valid_targets_mean": 28861.2, + "valid_targets_min": 21387 + }, + { + "epoch": 5.069370330843117, + "grad_norm": 0.29660186802360977, + "learning_rate": 7.125854583712615e-05, + "loss": 0.4096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4084359407424927, + "step": 4750, + "valid_targets_mean": 29104.8, + "valid_targets_min": 22716 + }, + { + "epoch": 5.074706510138741, + "grad_norm": 0.3010971793356668, + "learning_rate": 7.103560014842308e-05, + "loss": 0.4122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4073614180088043, + "step": 4755, + "valid_targets_mean": 28881.0, + "valid_targets_min": 22729 + }, + { + "epoch": 5.080042689434365, + "grad_norm": 0.2862444691931529, + "learning_rate": 7.08128115001293e-05, + "loss": 0.4112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4098625183105469, + "step": 4760, + "valid_targets_mean": 28961.6, + "valid_targets_min": 19299 + }, + { + "epoch": 5.08537886872999, + "grad_norm": 0.25020308493368776, + "learning_rate": 7.059018110016973e-05, + "loss": 0.4112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4086630642414093, + "step": 4765, + "valid_targets_mean": 29057.1, + "valid_targets_min": 22094 + }, + { + "epoch": 5.090715048025614, + "grad_norm": 0.27847527159225727, + "learning_rate": 7.036771015561121e-05, + "loss": 0.4126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4140252470970154, + "step": 4770, + "valid_targets_mean": 28867.9, + "valid_targets_min": 21388 + }, + { + "epoch": 5.096051227321238, + "grad_norm": 0.3011903014699775, + "learning_rate": 7.01453998726562e-05, + "loss": 0.4094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.411713182926178, + "step": 4775, + "valid_targets_mean": 28863.7, + "valid_targets_min": 20348 + }, + { + "epoch": 5.101387406616863, + "grad_norm": 0.30730661666666664, + "learning_rate": 6.992325145663598e-05, + "loss": 0.4123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4111214280128479, + "step": 4780, + "valid_targets_mean": 28753.7, + "valid_targets_min": 22883 + }, + { + "epoch": 5.106723585912487, + "grad_norm": 0.32196774097231146, + "learning_rate": 6.970126611200427e-05, + "loss": 0.4105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41407275199890137, + "step": 4785, + "valid_targets_mean": 28852.9, + "valid_targets_min": 20224 + }, + { + "epoch": 5.112059765208111, + "grad_norm": 0.30632694717862874, + "learning_rate": 6.947944504233056e-05, + "loss": 0.4126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41604679822921753, + "step": 4790, + "valid_targets_mean": 28705.6, + "valid_targets_min": 19229 + }, + { + "epoch": 5.117395944503735, + "grad_norm": 0.253831090174983, + "learning_rate": 6.925778945029372e-05, + "loss": 0.4113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4120824933052063, + "step": 4795, + "valid_targets_mean": 28878.8, + "valid_targets_min": 22820 + }, + { + "epoch": 5.122732123799359, + "grad_norm": 0.3456780929737466, + "learning_rate": 6.903630053767547e-05, + "loss": 0.4123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4113757014274597, + "step": 4800, + "valid_targets_mean": 28891.6, + "valid_targets_min": 19016 + }, + { + "epoch": 5.128068303094984, + "grad_norm": 0.3165719394775509, + "learning_rate": 6.881497950535372e-05, + "loss": 0.4122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41525954008102417, + "step": 4805, + "valid_targets_mean": 28823.5, + "valid_targets_min": 19620 + }, + { + "epoch": 5.133404482390608, + "grad_norm": 0.2661781179345722, + "learning_rate": 6.859382755329623e-05, + "loss": 0.4118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4119814336299896, + "step": 4810, + "valid_targets_mean": 28951.9, + "valid_targets_min": 21979 + }, + { + "epoch": 5.138740661686232, + "grad_norm": 0.3053995831655596, + "learning_rate": 6.837284588055401e-05, + "loss": 0.4135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41530489921569824, + "step": 4815, + "valid_targets_mean": 28814.4, + "valid_targets_min": 22303 + }, + { + "epoch": 5.144076840981857, + "grad_norm": 0.2985414542902706, + "learning_rate": 6.81520356852549e-05, + "loss": 0.4103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4107014536857605, + "step": 4820, + "valid_targets_mean": 28636.6, + "valid_targets_min": 19438 + }, + { + "epoch": 5.149413020277481, + "grad_norm": 0.27395231323939345, + "learning_rate": 6.793139816459688e-05, + "loss": 0.4116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41278764605522156, + "step": 4825, + "valid_targets_mean": 28825.8, + "valid_targets_min": 22834 + }, + { + "epoch": 5.154749199573105, + "grad_norm": 0.2815876337160247, + "learning_rate": 6.771093451484184e-05, + "loss": 0.411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4093128740787506, + "step": 4830, + "valid_targets_mean": 28861.8, + "valid_targets_min": 20543 + }, + { + "epoch": 5.16008537886873, + "grad_norm": 0.2577392190350307, + "learning_rate": 6.749064593130896e-05, + "loss": 0.4116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4116304814815521, + "step": 4835, + "valid_targets_mean": 28712.0, + "valid_targets_min": 21420 + }, + { + "epoch": 5.165421558164354, + "grad_norm": 0.32451201108936994, + "learning_rate": 6.727053360836813e-05, + "loss": 0.4125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4091717004776001, + "step": 4840, + "valid_targets_mean": 28862.9, + "valid_targets_min": 19087 + }, + { + "epoch": 5.1707577374599785, + "grad_norm": 0.26628464480777275, + "learning_rate": 6.705059873943375e-05, + "loss": 0.4103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41106921434402466, + "step": 4845, + "valid_targets_mean": 28710.1, + "valid_targets_min": 16993 + }, + { + "epoch": 5.176093916755603, + "grad_norm": 0.2698354108808259, + "learning_rate": 6.683084251695794e-05, + "loss": 0.4085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4063991904258728, + "step": 4850, + "valid_targets_mean": 28867.3, + "valid_targets_min": 22244 + }, + { + "epoch": 5.181430096051227, + "grad_norm": 0.28447317240929537, + "learning_rate": 6.66112661324243e-05, + "loss": 0.4125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4150131046772003, + "step": 4855, + "valid_targets_mean": 28886.2, + "valid_targets_min": 18504 + }, + { + "epoch": 5.1867662753468515, + "grad_norm": 0.3206606983392642, + "learning_rate": 6.639187077634141e-05, + "loss": 0.4113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41379764676094055, + "step": 4860, + "valid_targets_mean": 28757.9, + "valid_targets_min": 21961 + }, + { + "epoch": 5.192102454642476, + "grad_norm": 0.2944187342507382, + "learning_rate": 6.617265763823623e-05, + "loss": 0.4118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.410386860370636, + "step": 4865, + "valid_targets_mean": 28685.1, + "valid_targets_min": 21053 + }, + { + "epoch": 5.1974386339381, + "grad_norm": 0.31442256997337714, + "learning_rate": 6.59536279066479e-05, + "loss": 0.4115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40691661834716797, + "step": 4870, + "valid_targets_mean": 28880.3, + "valid_targets_min": 20361 + }, + { + "epoch": 5.202774813233725, + "grad_norm": 0.28243262362895716, + "learning_rate": 6.573478276912102e-05, + "loss": 0.4098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40908265113830566, + "step": 4875, + "valid_targets_mean": 28886.4, + "valid_targets_min": 22533 + }, + { + "epoch": 5.208110992529349, + "grad_norm": 0.29137903655336894, + "learning_rate": 6.551612341219951e-05, + "loss": 0.4104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4124106168746948, + "step": 4880, + "valid_targets_mean": 28826.8, + "valid_targets_min": 21510 + }, + { + "epoch": 5.213447171824973, + "grad_norm": 0.3062013407968183, + "learning_rate": 6.529765102141987e-05, + "loss": 0.4085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4086436927318573, + "step": 4885, + "valid_targets_mean": 28920.0, + "valid_targets_min": 21965 + }, + { + "epoch": 5.218783351120598, + "grad_norm": 0.30435542062456195, + "learning_rate": 6.507936678130493e-05, + "loss": 0.4107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4137675166130066, + "step": 4890, + "valid_targets_mean": 28880.1, + "valid_targets_min": 22355 + }, + { + "epoch": 5.224119530416222, + "grad_norm": 0.3029459152474519, + "learning_rate": 6.486127187535749e-05, + "loss": 0.4107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41013699769973755, + "step": 4895, + "valid_targets_mean": 28795.0, + "valid_targets_min": 19807 + }, + { + "epoch": 5.229455709711846, + "grad_norm": 0.27318022386491375, + "learning_rate": 6.464336748605368e-05, + "loss": 0.4121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4109240770339966, + "step": 4900, + "valid_targets_mean": 28784.6, + "valid_targets_min": 22323 + }, + { + "epoch": 5.234791889007471, + "grad_norm": 0.28034368378085794, + "learning_rate": 6.44256547948368e-05, + "loss": 0.4103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4095957577228546, + "step": 4905, + "valid_targets_mean": 28739.1, + "valid_targets_min": 22580 + }, + { + "epoch": 5.240128068303095, + "grad_norm": 0.28199830261418546, + "learning_rate": 6.420813498211071e-05, + "loss": 0.4094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4090818464756012, + "step": 4910, + "valid_targets_mean": 28894.7, + "valid_targets_min": 18300 + }, + { + "epoch": 5.245464247598719, + "grad_norm": 0.27836654922584375, + "learning_rate": 6.399080922723352e-05, + "loss": 0.4098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41042056679725647, + "step": 4915, + "valid_targets_mean": 28878.4, + "valid_targets_min": 21215 + }, + { + "epoch": 5.250800426894344, + "grad_norm": 0.24496887304740644, + "learning_rate": 6.377367870851124e-05, + "loss": 0.4112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41321632266044617, + "step": 4920, + "valid_targets_mean": 28918.7, + "valid_targets_min": 22650 + }, + { + "epoch": 5.256136606189968, + "grad_norm": 0.26396913584127146, + "learning_rate": 6.355674460319128e-05, + "loss": 0.4091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4113008975982666, + "step": 4925, + "valid_targets_mean": 28809.0, + "valid_targets_min": 21227 + }, + { + "epoch": 5.2614727854855925, + "grad_norm": 0.3216499292759843, + "learning_rate": 6.334000808745617e-05, + "loss": 0.41, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40966159105300903, + "step": 4930, + "valid_targets_mean": 29048.1, + "valid_targets_min": 21325 + }, + { + "epoch": 5.266808964781217, + "grad_norm": 0.2518698461231176, + "learning_rate": 6.312347033641712e-05, + "loss": 0.409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40684419870376587, + "step": 4935, + "valid_targets_mean": 29002.1, + "valid_targets_min": 23299 + }, + { + "epoch": 5.272145144076841, + "grad_norm": 0.26439597268229703, + "learning_rate": 6.290713252410772e-05, + "loss": 0.4097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4131563901901245, + "step": 4940, + "valid_targets_mean": 28740.8, + "valid_targets_min": 20226 + }, + { + "epoch": 5.277481323372466, + "grad_norm": 0.2482891922972573, + "learning_rate": 6.269099582347743e-05, + "loss": 0.4096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4077172875404358, + "step": 4945, + "valid_targets_mean": 28777.4, + "valid_targets_min": 20923 + }, + { + "epoch": 5.28281750266809, + "grad_norm": 0.2901076069620523, + "learning_rate": 6.247506140638535e-05, + "loss": 0.4104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41215550899505615, + "step": 4950, + "valid_targets_mean": 28667.8, + "valid_targets_min": 17866 + }, + { + "epoch": 5.288153681963714, + "grad_norm": 0.2588091918678108, + "learning_rate": 6.225933044359389e-05, + "loss": 0.4117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40798309445381165, + "step": 4955, + "valid_targets_mean": 28690.2, + "valid_targets_min": 19451 + }, + { + "epoch": 5.293489861259339, + "grad_norm": 0.32159144819204216, + "learning_rate": 6.204380410476227e-05, + "loss": 0.4103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4119293689727783, + "step": 4960, + "valid_targets_mean": 28846.5, + "valid_targets_min": 15600 + }, + { + "epoch": 5.298826040554963, + "grad_norm": 0.24255140450372784, + "learning_rate": 6.182848355844035e-05, + "loss": 0.4094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4105904996395111, + "step": 4965, + "valid_targets_mean": 28667.0, + "valid_targets_min": 21401 + }, + { + "epoch": 5.304162219850587, + "grad_norm": 0.27448018537267005, + "learning_rate": 6.161336997206222e-05, + "loss": 0.4109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4086799621582031, + "step": 4970, + "valid_targets_mean": 28726.4, + "valid_targets_min": 20804 + }, + { + "epoch": 5.309498399146212, + "grad_norm": 0.2566682976560549, + "learning_rate": 6.13984645119397e-05, + "loss": 0.4096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40948301553726196, + "step": 4975, + "valid_targets_mean": 28890.5, + "valid_targets_min": 19729 + }, + { + "epoch": 5.314834578441836, + "grad_norm": 0.2594204239198475, + "learning_rate": 6.118376834325645e-05, + "loss": 0.4112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4136618971824646, + "step": 4980, + "valid_targets_mean": 28906.8, + "valid_targets_min": 22746 + }, + { + "epoch": 5.3201707577374595, + "grad_norm": 0.2925635711676992, + "learning_rate": 6.0969282630061165e-05, + "loss": 0.4104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4110929071903229, + "step": 4985, + "valid_targets_mean": 28713.1, + "valid_targets_min": 22006 + }, + { + "epoch": 5.325506937033085, + "grad_norm": 0.3313501113694902, + "learning_rate": 6.0755008535261626e-05, + "loss": 0.4125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4129217863082886, + "step": 4990, + "valid_targets_mean": 28719.1, + "valid_targets_min": 19168 + }, + { + "epoch": 5.330843116328708, + "grad_norm": 0.2798300156894228, + "learning_rate": 6.054094722061812e-05, + "loss": 0.411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4137015640735626, + "step": 4995, + "valid_targets_mean": 28776.3, + "valid_targets_min": 22377 + }, + { + "epoch": 5.336179295624333, + "grad_norm": 0.26481226416695713, + "learning_rate": 6.032709984673739e-05, + "loss": 0.409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40870779752731323, + "step": 5000, + "valid_targets_mean": 28733.4, + "valid_targets_min": 16381 + }, + { + "epoch": 5.341515474919957, + "grad_norm": 0.2574571665394944, + "learning_rate": 6.0113467573066174e-05, + "loss": 0.4068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4053928852081299, + "step": 5005, + "valid_targets_mean": 28782.5, + "valid_targets_min": 21917 + }, + { + "epoch": 5.346851654215581, + "grad_norm": 0.2719571419049441, + "learning_rate": 5.990005155788495e-05, + "loss": 0.4106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41064536571502686, + "step": 5010, + "valid_targets_mean": 28818.9, + "valid_targets_min": 20918 + }, + { + "epoch": 5.352187833511206, + "grad_norm": 0.26204971312606534, + "learning_rate": 5.9686852958301765e-05, + "loss": 0.4085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41017162799835205, + "step": 5015, + "valid_targets_mean": 28822.9, + "valid_targets_min": 21199 + }, + { + "epoch": 5.35752401280683, + "grad_norm": 0.26277915538565405, + "learning_rate": 5.9473872930245745e-05, + "loss": 0.409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.407803475856781, + "step": 5020, + "valid_targets_mean": 28851.6, + "valid_targets_min": 20387 + }, + { + "epoch": 5.362860192102454, + "grad_norm": 0.2532508057206155, + "learning_rate": 5.9261112628461056e-05, + "loss": 0.4078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4040972590446472, + "step": 5025, + "valid_targets_mean": 28807.4, + "valid_targets_min": 20681 + }, + { + "epoch": 5.368196371398079, + "grad_norm": 0.26826146816164903, + "learning_rate": 5.9048573206500536e-05, + "loss": 0.4089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4086844027042389, + "step": 5030, + "valid_targets_mean": 28767.2, + "valid_targets_min": 17033 + }, + { + "epoch": 5.373532550693703, + "grad_norm": 0.2732203900470291, + "learning_rate": 5.883625581671934e-05, + "loss": 0.4076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40953680872917175, + "step": 5035, + "valid_targets_mean": 28861.1, + "valid_targets_min": 20183 + }, + { + "epoch": 5.378868729989327, + "grad_norm": 0.26014816155601717, + "learning_rate": 5.8624161610268956e-05, + "loss": 0.4089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40767329931259155, + "step": 5040, + "valid_targets_mean": 28843.2, + "valid_targets_min": 18915 + }, + { + "epoch": 5.384204909284952, + "grad_norm": 0.22818084752138812, + "learning_rate": 5.8412291737090664e-05, + "loss": 0.4068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40318650007247925, + "step": 5045, + "valid_targets_mean": 28832.0, + "valid_targets_min": 22102 + }, + { + "epoch": 5.389541088580576, + "grad_norm": 0.27327260028041966, + "learning_rate": 5.820064734590959e-05, + "loss": 0.408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4099733233451843, + "step": 5050, + "valid_targets_mean": 28704.1, + "valid_targets_min": 19957 + }, + { + "epoch": 5.3948772678762005, + "grad_norm": 0.28859428097531076, + "learning_rate": 5.798922958422819e-05, + "loss": 0.4084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4103071987628937, + "step": 5055, + "valid_targets_mean": 28830.3, + "valid_targets_min": 21361 + }, + { + "epoch": 5.400213447171825, + "grad_norm": 0.22978650256782315, + "learning_rate": 5.777803959832029e-05, + "loss": 0.4092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4068617820739746, + "step": 5060, + "valid_targets_mean": 28820.3, + "valid_targets_min": 22322 + }, + { + "epoch": 5.405549626467449, + "grad_norm": 0.27127959486835485, + "learning_rate": 5.756707853322466e-05, + "loss": 0.4084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40688830614089966, + "step": 5065, + "valid_targets_mean": 28791.6, + "valid_targets_min": 21860 + }, + { + "epoch": 5.4108858057630735, + "grad_norm": 0.2506525033173379, + "learning_rate": 5.7356347532738906e-05, + "loss": 0.4087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41265010833740234, + "step": 5070, + "valid_targets_mean": 28704.1, + "valid_targets_min": 20010 + }, + { + "epoch": 5.416221985058698, + "grad_norm": 0.2751486277336242, + "learning_rate": 5.714584773941336e-05, + "loss": 0.4081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41021281480789185, + "step": 5075, + "valid_targets_mean": 28805.6, + "valid_targets_min": 20428 + }, + { + "epoch": 5.421558164354322, + "grad_norm": 0.24266350069275283, + "learning_rate": 5.693558029454458e-05, + "loss": 0.4098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4110463857650757, + "step": 5080, + "valid_targets_mean": 28593.4, + "valid_targets_min": 20891 + }, + { + "epoch": 5.426894343649947, + "grad_norm": 0.27938597765654943, + "learning_rate": 5.6725546338169675e-05, + "loss": 0.4084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40841275453567505, + "step": 5085, + "valid_targets_mean": 29122.6, + "valid_targets_min": 19030 + }, + { + "epoch": 5.432230522945571, + "grad_norm": 0.2895610586627878, + "learning_rate": 5.651574700905955e-05, + "loss": 0.4075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40801024436950684, + "step": 5090, + "valid_targets_mean": 28741.1, + "valid_targets_min": 21502 + }, + { + "epoch": 5.437566702241195, + "grad_norm": 0.2427061435506355, + "learning_rate": 5.6306183444713066e-05, + "loss": 0.407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4059031903743744, + "step": 5095, + "valid_targets_mean": 28847.6, + "valid_targets_min": 22709 + }, + { + "epoch": 5.44290288153682, + "grad_norm": 0.2779281039684238, + "learning_rate": 5.609685678135089e-05, + "loss": 0.4078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4095008969306946, + "step": 5100, + "valid_targets_mean": 28784.6, + "valid_targets_min": 19776 + }, + { + "epoch": 5.448239060832444, + "grad_norm": 0.26779235200255996, + "learning_rate": 5.588776815390913e-05, + "loss": 0.4082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40894991159439087, + "step": 5105, + "valid_targets_mean": 28831.8, + "valid_targets_min": 23168 + }, + { + "epoch": 5.453575240128068, + "grad_norm": 0.2838087294033303, + "learning_rate": 5.5678918696033355e-05, + "loss": 0.4088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4082152843475342, + "step": 5110, + "valid_targets_mean": 28729.3, + "valid_targets_min": 18829 + }, + { + "epoch": 5.458911419423693, + "grad_norm": 0.2305536584413741, + "learning_rate": 5.5470309540072443e-05, + "loss": 0.406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.402679443359375, + "step": 5115, + "valid_targets_mean": 28995.2, + "valid_targets_min": 21516 + }, + { + "epoch": 5.464247598719317, + "grad_norm": 0.22816329039508723, + "learning_rate": 5.526194181707236e-05, + "loss": 0.408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4087204039096832, + "step": 5120, + "valid_targets_mean": 28795.0, + "valid_targets_min": 19489 + }, + { + "epoch": 5.4695837780149414, + "grad_norm": 0.24529677775400807, + "learning_rate": 5.505381665677001e-05, + "loss": 0.4092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4101272225379944, + "step": 5125, + "valid_targets_mean": 28636.4, + "valid_targets_min": 16244 + }, + { + "epoch": 5.474919957310566, + "grad_norm": 0.28571712158040247, + "learning_rate": 5.484593518758716e-05, + "loss": 0.41, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4107014238834381, + "step": 5130, + "valid_targets_mean": 28694.7, + "valid_targets_min": 22274 + }, + { + "epoch": 5.48025613660619, + "grad_norm": 0.24596211486579148, + "learning_rate": 5.4638298536624445e-05, + "loss": 0.409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40564632415771484, + "step": 5135, + "valid_targets_mean": 28841.2, + "valid_targets_min": 21563 + }, + { + "epoch": 5.4855923159018145, + "grad_norm": 0.26465991092206825, + "learning_rate": 5.443090782965496e-05, + "loss": 0.4092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4095415472984314, + "step": 5140, + "valid_targets_mean": 28906.0, + "valid_targets_min": 21832 + }, + { + "epoch": 5.490928495197439, + "grad_norm": 0.22267519414163728, + "learning_rate": 5.422376419111848e-05, + "loss": 0.4079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40773892402648926, + "step": 5145, + "valid_targets_mean": 28854.4, + "valid_targets_min": 22257 + }, + { + "epoch": 5.496264674493063, + "grad_norm": 0.2808312854307953, + "learning_rate": 5.401686874411516e-05, + "loss": 0.4053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40362006425857544, + "step": 5150, + "valid_targets_mean": 28736.9, + "valid_targets_min": 18357 + }, + { + "epoch": 5.501600853788688, + "grad_norm": 0.28073780928417497, + "learning_rate": 5.381022261039944e-05, + "loss": 0.4089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4081991910934448, + "step": 5155, + "valid_targets_mean": 28902.0, + "valid_targets_min": 22967 + }, + { + "epoch": 5.506937033084312, + "grad_norm": 0.2631869993080492, + "learning_rate": 5.3603826910374153e-05, + "loss": 0.407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40714550018310547, + "step": 5160, + "valid_targets_mean": 28665.1, + "valid_targets_min": 21412 + }, + { + "epoch": 5.512273212379936, + "grad_norm": 0.2884581744505189, + "learning_rate": 5.3397682763084145e-05, + "loss": 0.4085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4094289541244507, + "step": 5165, + "valid_targets_mean": 28766.8, + "valid_targets_min": 19331 + }, + { + "epoch": 5.517609391675561, + "grad_norm": 0.2575293949311548, + "learning_rate": 5.319179128621053e-05, + "loss": 0.408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4074215888977051, + "step": 5170, + "valid_targets_mean": 28829.6, + "valid_targets_min": 20404 + }, + { + "epoch": 5.522945570971185, + "grad_norm": 0.2752574398921403, + "learning_rate": 5.2986153596064446e-05, + "loss": 0.4082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4028487801551819, + "step": 5175, + "valid_targets_mean": 28834.2, + "valid_targets_min": 22315 + }, + { + "epoch": 5.528281750266809, + "grad_norm": 0.26181590292808854, + "learning_rate": 5.278077080758106e-05, + "loss": 0.4063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40661585330963135, + "step": 5180, + "valid_targets_mean": 28683.8, + "valid_targets_min": 19502 + }, + { + "epoch": 5.533617929562434, + "grad_norm": 0.30026428350022577, + "learning_rate": 5.257564403431341e-05, + "loss": 0.4095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41059237718582153, + "step": 5185, + "valid_targets_mean": 28661.4, + "valid_targets_min": 20971 + }, + { + "epoch": 5.538954108858057, + "grad_norm": 0.33290674109160145, + "learning_rate": 5.23707743884265e-05, + "loss": 0.4076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40834152698516846, + "step": 5190, + "valid_targets_mean": 28679.2, + "valid_targets_min": 18819 + }, + { + "epoch": 5.544290288153682, + "grad_norm": 0.2451716687040915, + "learning_rate": 5.216616298069132e-05, + "loss": 0.4059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40883660316467285, + "step": 5195, + "valid_targets_mean": 28750.6, + "valid_targets_min": 21745 + }, + { + "epoch": 5.549626467449306, + "grad_norm": 0.2518337780712598, + "learning_rate": 5.196181092047858e-05, + "loss": 0.4065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4079820513725281, + "step": 5200, + "valid_targets_mean": 28827.4, + "valid_targets_min": 19230 + }, + { + "epoch": 5.554962646744931, + "grad_norm": 0.23555228050632473, + "learning_rate": 5.1757719315753e-05, + "loss": 0.4092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41063809394836426, + "step": 5205, + "valid_targets_mean": 28860.7, + "valid_targets_min": 20761 + }, + { + "epoch": 5.560298826040555, + "grad_norm": 0.23131626823494578, + "learning_rate": 5.155388927306707e-05, + "loss": 0.4081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40856459736824036, + "step": 5210, + "valid_targets_mean": 28866.7, + "valid_targets_min": 21514 + }, + { + "epoch": 5.565635005336179, + "grad_norm": 0.24583588381945762, + "learning_rate": 5.13503218975551e-05, + "loss": 0.4072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4081979990005493, + "step": 5215, + "valid_targets_mean": 28815.9, + "valid_targets_min": 22727 + }, + { + "epoch": 5.570971184631803, + "grad_norm": 0.26323488215272317, + "learning_rate": 5.1147018292927384e-05, + "loss": 0.4071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4045487344264984, + "step": 5220, + "valid_targets_mean": 28845.8, + "valid_targets_min": 22600 + }, + { + "epoch": 5.576307363927428, + "grad_norm": 0.2617473214116886, + "learning_rate": 5.094397956146391e-05, + "loss": 0.4079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40857842564582825, + "step": 5225, + "valid_targets_mean": 28835.9, + "valid_targets_min": 22871 + }, + { + "epoch": 5.581643543223052, + "grad_norm": 0.230237887341813, + "learning_rate": 5.074120680400877e-05, + "loss": 0.4058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4065649211406708, + "step": 5230, + "valid_targets_mean": 28784.2, + "valid_targets_min": 22150 + }, + { + "epoch": 5.586979722518676, + "grad_norm": 0.23212744439681787, + "learning_rate": 5.053870111996379e-05, + "loss": 0.4092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4103742241859436, + "step": 5235, + "valid_targets_mean": 28789.1, + "valid_targets_min": 22110 + }, + { + "epoch": 5.592315901814301, + "grad_norm": 0.2231287372517266, + "learning_rate": 5.0336463607282915e-05, + "loss": 0.4086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.411297082901001, + "step": 5240, + "valid_targets_mean": 28775.0, + "valid_targets_min": 21938 + }, + { + "epoch": 5.597652081109925, + "grad_norm": 0.2686855989256489, + "learning_rate": 5.013449536246607e-05, + "loss": 0.4071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.411526620388031, + "step": 5245, + "valid_targets_mean": 28883.7, + "valid_targets_min": 16759 + }, + { + "epoch": 5.602988260405549, + "grad_norm": 0.26466533071442333, + "learning_rate": 4.9932797480553174e-05, + "loss": 0.4054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40664201974868774, + "step": 5250, + "valid_targets_mean": 28914.4, + "valid_targets_min": 23391 + }, + { + "epoch": 5.608324439701174, + "grad_norm": 0.21338154387922556, + "learning_rate": 4.973137105511842e-05, + "loss": 0.4084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40618956089019775, + "step": 5255, + "valid_targets_mean": 28953.4, + "valid_targets_min": 22290 + }, + { + "epoch": 5.613660618996798, + "grad_norm": 0.22845809094870723, + "learning_rate": 4.9530217178264035e-05, + "loss": 0.4078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4072515368461609, + "step": 5260, + "valid_targets_mean": 28862.8, + "valid_targets_min": 20440 + }, + { + "epoch": 5.6189967982924225, + "grad_norm": 0.23349213346807032, + "learning_rate": 4.932933694061469e-05, + "loss": 0.4081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4063612222671509, + "step": 5265, + "valid_targets_mean": 28641.9, + "valid_targets_min": 13903 + }, + { + "epoch": 5.624332977588047, + "grad_norm": 0.2624107603135616, + "learning_rate": 4.912873143131137e-05, + "loss": 0.4079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40512439608573914, + "step": 5270, + "valid_targets_mean": 29003.0, + "valid_targets_min": 20872 + }, + { + "epoch": 5.629669156883671, + "grad_norm": 0.2226041520979999, + "learning_rate": 4.892840173800545e-05, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4093719720840454, + "step": 5275, + "valid_targets_mean": 29018.0, + "valid_targets_min": 18214 + }, + { + "epoch": 5.6350053361792956, + "grad_norm": 0.25335082989920943, + "learning_rate": 4.872834894685303e-05, + "loss": 0.4079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4051392674446106, + "step": 5280, + "valid_targets_mean": 28871.5, + "valid_targets_min": 21415 + }, + { + "epoch": 5.64034151547492, + "grad_norm": 0.26416237862885106, + "learning_rate": 4.852857414250871e-05, + "loss": 0.4072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4078800678253174, + "step": 5285, + "valid_targets_mean": 28953.2, + "valid_targets_min": 23044 + }, + { + "epoch": 5.645677694770544, + "grad_norm": 0.2594465825907819, + "learning_rate": 4.832907840812007e-05, + "loss": 0.4094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4099881052970886, + "step": 5290, + "valid_targets_mean": 29002.3, + "valid_targets_min": 23780 + }, + { + "epoch": 5.651013874066169, + "grad_norm": 0.25857051667325454, + "learning_rate": 4.812986282532144e-05, + "loss": 0.4074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40799394249916077, + "step": 5295, + "valid_targets_mean": 28782.3, + "valid_targets_min": 21471 + }, + { + "epoch": 5.656350053361793, + "grad_norm": 0.23165297682274635, + "learning_rate": 4.7930928474228345e-05, + "loss": 0.4072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4110855460166931, + "step": 5300, + "valid_targets_mean": 28889.3, + "valid_targets_min": 19959 + }, + { + "epoch": 5.661686232657417, + "grad_norm": 0.21714603686390757, + "learning_rate": 4.7732276433431514e-05, + "loss": 0.4079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4074994623661041, + "step": 5305, + "valid_targets_mean": 28760.8, + "valid_targets_min": 17978 + }, + { + "epoch": 5.667022411953042, + "grad_norm": 0.25121345612741014, + "learning_rate": 4.753390777999091e-05, + "loss": 0.4075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40740853548049927, + "step": 5310, + "valid_targets_mean": 28903.8, + "valid_targets_min": 18702 + }, + { + "epoch": 5.672358591248666, + "grad_norm": 0.24076860265064987, + "learning_rate": 4.733582358943016e-05, + "loss": 0.4086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4073009192943573, + "step": 5315, + "valid_targets_mean": 28919.6, + "valid_targets_min": 18070 + }, + { + "epoch": 5.67769477054429, + "grad_norm": 0.24410116917299898, + "learning_rate": 4.7138024935730464e-05, + "loss": 0.407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40609365701675415, + "step": 5320, + "valid_targets_mean": 28835.2, + "valid_targets_min": 22006 + }, + { + "epoch": 5.683030949839915, + "grad_norm": 0.22406370130790632, + "learning_rate": 4.6940512891324986e-05, + "loss": 0.4076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40902549028396606, + "step": 5325, + "valid_targets_mean": 28876.1, + "valid_targets_min": 17173 + }, + { + "epoch": 5.688367129135539, + "grad_norm": 0.2220167506558589, + "learning_rate": 4.674328852709283e-05, + "loss": 0.4062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4030117988586426, + "step": 5330, + "valid_targets_mean": 28941.9, + "valid_targets_min": 23033 + }, + { + "epoch": 5.6937033084311635, + "grad_norm": 0.2461889251597465, + "learning_rate": 4.654635291235343e-05, + "loss": 0.4064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40485891699790955, + "step": 5335, + "valid_targets_mean": 28843.3, + "valid_targets_min": 22189 + }, + { + "epoch": 5.699039487726788, + "grad_norm": 0.2868230592419842, + "learning_rate": 4.634970711486068e-05, + "loss": 0.4068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4057909846305847, + "step": 5340, + "valid_targets_mean": 28854.3, + "valid_targets_min": 22418 + }, + { + "epoch": 5.704375667022412, + "grad_norm": 0.24789472475723118, + "learning_rate": 4.615335220079703e-05, + "loss": 0.4055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4078221321105957, + "step": 5345, + "valid_targets_mean": 28942.5, + "valid_targets_min": 21262 + }, + { + "epoch": 5.7097118463180365, + "grad_norm": 0.25318073622340376, + "learning_rate": 4.595728923476792e-05, + "loss": 0.4069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40469861030578613, + "step": 5350, + "valid_targets_mean": 28845.2, + "valid_targets_min": 19004 + }, + { + "epoch": 5.715048025613661, + "grad_norm": 0.2580212397753865, + "learning_rate": 4.576151927979577e-05, + "loss": 0.4072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4108935594558716, + "step": 5355, + "valid_targets_mean": 28843.6, + "valid_targets_min": 19456 + }, + { + "epoch": 5.720384204909285, + "grad_norm": 0.22310778312178198, + "learning_rate": 4.556604339731445e-05, + "loss": 0.4078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41052567958831787, + "step": 5360, + "valid_targets_mean": 28791.6, + "valid_targets_min": 18590 + }, + { + "epoch": 5.72572038420491, + "grad_norm": 0.2681032818849847, + "learning_rate": 4.537086264716341e-05, + "loss": 0.4085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4068934917449951, + "step": 5365, + "valid_targets_mean": 28867.4, + "valid_targets_min": 23073 + }, + { + "epoch": 5.731056563500534, + "grad_norm": 0.23148769627141919, + "learning_rate": 4.517597808758183e-05, + "loss": 0.4054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4052630662918091, + "step": 5370, + "valid_targets_mean": 28865.1, + "valid_targets_min": 21222 + }, + { + "epoch": 5.736392742796158, + "grad_norm": 0.22042817501892076, + "learning_rate": 4.498139077520314e-05, + "loss": 0.4071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4036562442779541, + "step": 5375, + "valid_targets_mean": 28880.2, + "valid_targets_min": 19297 + }, + { + "epoch": 5.741728922091783, + "grad_norm": 0.23595064942007865, + "learning_rate": 4.478710176504898e-05, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4078393876552582, + "step": 5380, + "valid_targets_mean": 28741.1, + "valid_targets_min": 19445 + }, + { + "epoch": 5.747065101387407, + "grad_norm": 0.2356224677462118, + "learning_rate": 4.459311211052382e-05, + "loss": 0.4067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40955275297164917, + "step": 5385, + "valid_targets_mean": 28913.4, + "valid_targets_min": 22165 + }, + { + "epoch": 5.752401280683031, + "grad_norm": 0.23613025476452645, + "learning_rate": 4.43994228634089e-05, + "loss": 0.4058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4061693251132965, + "step": 5390, + "valid_targets_mean": 28761.6, + "valid_targets_min": 16988 + }, + { + "epoch": 5.757737459978655, + "grad_norm": 0.2509453674002872, + "learning_rate": 4.4206035073856834e-05, + "loss": 0.4079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41114407777786255, + "step": 5395, + "valid_targets_mean": 28747.3, + "valid_targets_min": 17862 + }, + { + "epoch": 5.76307363927428, + "grad_norm": 0.21299150498723463, + "learning_rate": 4.4012949790385736e-05, + "loss": 0.4044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40109744668006897, + "step": 5400, + "valid_targets_mean": 28773.6, + "valid_targets_min": 19732 + }, + { + "epoch": 5.7684098185699035, + "grad_norm": 0.24642753596991596, + "learning_rate": 4.382016805987351e-05, + "loss": 0.407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4072338044643402, + "step": 5405, + "valid_targets_mean": 28985.3, + "valid_targets_min": 21313 + }, + { + "epoch": 5.773745997865529, + "grad_norm": 0.22244687905504107, + "learning_rate": 4.362769092755237e-05, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40808022022247314, + "step": 5410, + "valid_targets_mean": 28903.3, + "valid_targets_min": 18221 + }, + { + "epoch": 5.779082177161152, + "grad_norm": 0.2634528950267717, + "learning_rate": 4.343551943700291e-05, + "loss": 0.4065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4094908833503723, + "step": 5415, + "valid_targets_mean": 28849.7, + "valid_targets_min": 20021 + }, + { + "epoch": 5.784418356456777, + "grad_norm": 0.21314653878241394, + "learning_rate": 4.324365463014871e-05, + "loss": 0.4065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4070976674556732, + "step": 5420, + "valid_targets_mean": 28839.5, + "valid_targets_min": 21489 + }, + { + "epoch": 5.789754535752401, + "grad_norm": 0.2333881981036085, + "learning_rate": 4.3052097547250504e-05, + "loss": 0.4068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40807241201400757, + "step": 5425, + "valid_targets_mean": 28919.2, + "valid_targets_min": 21215 + }, + { + "epoch": 5.795090715048025, + "grad_norm": 0.22877396245428053, + "learning_rate": 4.2860849226900545e-05, + "loss": 0.4052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40419960021972656, + "step": 5430, + "valid_targets_mean": 28771.1, + "valid_targets_min": 21422 + }, + { + "epoch": 5.80042689434365, + "grad_norm": 0.25836705289249745, + "learning_rate": 4.2669910706017124e-05, + "loss": 0.4066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4077407121658325, + "step": 5435, + "valid_targets_mean": 28994.6, + "valid_targets_min": 22465 + }, + { + "epoch": 5.805763073639274, + "grad_norm": 0.22522683738726615, + "learning_rate": 4.2479283019838743e-05, + "loss": 0.4044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.401361882686615, + "step": 5440, + "valid_targets_mean": 28920.0, + "valid_targets_min": 18753 + }, + { + "epoch": 5.811099252934898, + "grad_norm": 0.22685710627505415, + "learning_rate": 4.228896720191873e-05, + "loss": 0.4063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4064394235610962, + "step": 5445, + "valid_targets_mean": 29037.0, + "valid_targets_min": 19954 + }, + { + "epoch": 5.816435432230523, + "grad_norm": 0.20142130133877767, + "learning_rate": 4.209896428411936e-05, + "loss": 0.4047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40420037508010864, + "step": 5450, + "valid_targets_mean": 28870.6, + "valid_targets_min": 18563 + }, + { + "epoch": 5.821771611526147, + "grad_norm": 0.20801333433848126, + "learning_rate": 4.190927529660653e-05, + "loss": 0.4056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4038311243057251, + "step": 5455, + "valid_targets_mean": 28757.4, + "valid_targets_min": 22298 + }, + { + "epoch": 5.827107790821771, + "grad_norm": 0.23230643362340275, + "learning_rate": 4.1719901267844086e-05, + "loss": 0.4053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40489107370376587, + "step": 5460, + "valid_targets_mean": 28917.4, + "valid_targets_min": 21333 + }, + { + "epoch": 5.832443970117396, + "grad_norm": 0.2164401832868788, + "learning_rate": 4.153084322458806e-05, + "loss": 0.4038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4066731631755829, + "step": 5465, + "valid_targets_mean": 28922.4, + "valid_targets_min": 23138 + }, + { + "epoch": 5.83778014941302, + "grad_norm": 0.21951841350335208, + "learning_rate": 4.134210219188141e-05, + "loss": 0.405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40590178966522217, + "step": 5470, + "valid_targets_mean": 28929.7, + "valid_targets_min": 22171 + }, + { + "epoch": 5.8431163287086445, + "grad_norm": 0.27913810326540894, + "learning_rate": 4.115367919304822e-05, + "loss": 0.4063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40679121017456055, + "step": 5475, + "valid_targets_mean": 28965.1, + "valid_targets_min": 19535 + }, + { + "epoch": 5.848452508004269, + "grad_norm": 0.2114164051428383, + "learning_rate": 4.0965575249688336e-05, + "loss": 0.4056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.403158962726593, + "step": 5480, + "valid_targets_mean": 28941.9, + "valid_targets_min": 20187 + }, + { + "epoch": 5.853788687299893, + "grad_norm": 0.256277757621636, + "learning_rate": 4.077779138167161e-05, + "loss": 0.4087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4057876169681549, + "step": 5485, + "valid_targets_mean": 28857.4, + "valid_targets_min": 20301 + }, + { + "epoch": 5.859124866595518, + "grad_norm": 0.21061324841836357, + "learning_rate": 4.059032860713261e-05, + "loss": 0.4055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40722543001174927, + "step": 5490, + "valid_targets_mean": 28856.1, + "valid_targets_min": 19253 + }, + { + "epoch": 5.864461045891142, + "grad_norm": 0.2518110545376062, + "learning_rate": 4.040318794246497e-05, + "loss": 0.4054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40780001878738403, + "step": 5495, + "valid_targets_mean": 28901.4, + "valid_targets_min": 21836 + }, + { + "epoch": 5.869797225186766, + "grad_norm": 0.2123751768365746, + "learning_rate": 4.0216370402315796e-05, + "loss": 0.4066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40632835030555725, + "step": 5500, + "valid_targets_mean": 28909.4, + "valid_targets_min": 20239 + }, + { + "epoch": 5.875133404482391, + "grad_norm": 0.21594602205244454, + "learning_rate": 4.002987699958042e-05, + "loss": 0.4053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40827763080596924, + "step": 5505, + "valid_targets_mean": 28695.7, + "valid_targets_min": 18285 + }, + { + "epoch": 5.880469583778015, + "grad_norm": 0.21362367358291603, + "learning_rate": 3.984370874539657e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4049973487854004, + "step": 5510, + "valid_targets_mean": 28791.7, + "valid_targets_min": 20225 + }, + { + "epoch": 5.885805763073639, + "grad_norm": 0.24153766266989785, + "learning_rate": 3.965786664913923e-05, + "loss": 0.4078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40766391158103943, + "step": 5515, + "valid_targets_mean": 28825.4, + "valid_targets_min": 22899 + }, + { + "epoch": 5.891141942369264, + "grad_norm": 0.23261290774247123, + "learning_rate": 3.947235171841497e-05, + "loss": 0.4055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40966296195983887, + "step": 5520, + "valid_targets_mean": 28849.6, + "valid_targets_min": 21756 + }, + { + "epoch": 5.896478121664888, + "grad_norm": 0.24097588461676905, + "learning_rate": 3.928716495905642e-05, + "loss": 0.4062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4064144492149353, + "step": 5525, + "valid_targets_mean": 29012.6, + "valid_targets_min": 20124 + }, + { + "epoch": 5.901814300960512, + "grad_norm": 0.21967799647204173, + "learning_rate": 3.9102307375117064e-05, + "loss": 0.4043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40806692838668823, + "step": 5530, + "valid_targets_mean": 28692.7, + "valid_targets_min": 21532 + }, + { + "epoch": 5.907150480256137, + "grad_norm": 0.21285101448089871, + "learning_rate": 3.891777996886551e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4044090211391449, + "step": 5535, + "valid_targets_mean": 28888.1, + "valid_targets_min": 22096 + }, + { + "epoch": 5.912486659551761, + "grad_norm": 0.2069155375792718, + "learning_rate": 3.873358374078033e-05, + "loss": 0.405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40243616700172424, + "step": 5540, + "valid_targets_mean": 28860.0, + "valid_targets_min": 18914 + }, + { + "epoch": 5.9178228388473855, + "grad_norm": 0.21705205284159565, + "learning_rate": 3.854971968954436e-05, + "loss": 0.4043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.401405394077301, + "step": 5545, + "valid_targets_mean": 28765.5, + "valid_targets_min": 17484 + }, + { + "epoch": 5.92315901814301, + "grad_norm": 0.2415797773522953, + "learning_rate": 3.836618881203953e-05, + "loss": 0.4072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40842103958129883, + "step": 5550, + "valid_targets_mean": 28809.4, + "valid_targets_min": 21450 + }, + { + "epoch": 5.928495197438634, + "grad_norm": 0.265004048605028, + "learning_rate": 3.8182992103341355e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40405726432800293, + "step": 5555, + "valid_targets_mean": 28818.4, + "valid_targets_min": 19185 + }, + { + "epoch": 5.9338313767342585, + "grad_norm": 0.21529807239273552, + "learning_rate": 3.800013055671343e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40286242961883545, + "step": 5560, + "valid_targets_mean": 28765.6, + "valid_targets_min": 18225 + }, + { + "epoch": 5.939167556029883, + "grad_norm": 0.21081684843965076, + "learning_rate": 3.781760516360232e-05, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4038587212562561, + "step": 5565, + "valid_targets_mean": 28933.4, + "valid_targets_min": 20988 + }, + { + "epoch": 5.944503735325507, + "grad_norm": 0.21843522151704156, + "learning_rate": 3.7635416913631874e-05, + "loss": 0.4051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40666764974594116, + "step": 5570, + "valid_targets_mean": 28948.6, + "valid_targets_min": 18118 + }, + { + "epoch": 5.949839914621132, + "grad_norm": 0.2038687063842565, + "learning_rate": 3.745356679459803e-05, + "loss": 0.4054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4015902280807495, + "step": 5575, + "valid_targets_mean": 28850.7, + "valid_targets_min": 22683 + }, + { + "epoch": 5.955176093916756, + "grad_norm": 0.24391497724401612, + "learning_rate": 3.727205579246349e-05, + "loss": 0.4059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40471071004867554, + "step": 5580, + "valid_targets_mean": 29030.9, + "valid_targets_min": 19158 + }, + { + "epoch": 5.96051227321238, + "grad_norm": 0.27555184336682575, + "learning_rate": 3.70908848913523e-05, + "loss": 0.4048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40131568908691406, + "step": 5585, + "valid_targets_mean": 28900.0, + "valid_targets_min": 21973 + }, + { + "epoch": 5.965848452508004, + "grad_norm": 0.21948801292048828, + "learning_rate": 3.6910055073544545e-05, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4104491174221039, + "step": 5590, + "valid_targets_mean": 28758.2, + "valid_targets_min": 16583 + }, + { + "epoch": 5.971184631803629, + "grad_norm": 0.21597893371804786, + "learning_rate": 3.672956731947093e-05, + "loss": 0.4047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4032149910926819, + "step": 5595, + "valid_targets_mean": 28712.1, + "valid_targets_min": 18488 + }, + { + "epoch": 5.9765208110992525, + "grad_norm": 0.23374622037111217, + "learning_rate": 3.6549422607707664e-05, + "loss": 0.4056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4101329445838928, + "step": 5600, + "valid_targets_mean": 28952.7, + "valid_targets_min": 23291 + }, + { + "epoch": 5.981856990394878, + "grad_norm": 0.225947924540093, + "learning_rate": 3.636962191497089e-05, + "loss": 0.4066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4063754081726074, + "step": 5605, + "valid_targets_mean": 28869.9, + "valid_targets_min": 22004 + }, + { + "epoch": 5.987193169690501, + "grad_norm": 0.24990410081389533, + "learning_rate": 3.6190166216111666e-05, + "loss": 0.4047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4043414890766144, + "step": 5610, + "valid_targets_mean": 29032.0, + "valid_targets_min": 20042 + }, + { + "epoch": 5.9925293489861255, + "grad_norm": 0.23553235177171936, + "learning_rate": 3.6011056484110525e-05, + "loss": 0.405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40443944931030273, + "step": 5615, + "valid_targets_mean": 29001.4, + "valid_targets_min": 19151 + }, + { + "epoch": 5.99786552828175, + "grad_norm": 0.2054776217712033, + "learning_rate": 3.5832293690072125e-05, + "loss": 0.4049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4076521396636963, + "step": 5620, + "valid_targets_mean": 28850.6, + "valid_targets_min": 22676 + }, + { + "epoch": 6.003201707577374, + "grad_norm": 0.22868682159600212, + "learning_rate": 3.565387880322022e-05, + "loss": 0.4039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4057372808456421, + "step": 5625, + "valid_targets_mean": 28841.2, + "valid_targets_min": 22815 + }, + { + "epoch": 6.008537886872999, + "grad_norm": 0.21873814325197052, + "learning_rate": 3.547581279089216e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4041280746459961, + "step": 5630, + "valid_targets_mean": 28757.2, + "valid_targets_min": 20584 + }, + { + "epoch": 6.013874066168623, + "grad_norm": 0.2037817066603688, + "learning_rate": 3.529809661853376e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4051162898540497, + "step": 5635, + "valid_targets_mean": 28794.5, + "valid_targets_min": 22709 + }, + { + "epoch": 6.019210245464247, + "grad_norm": 0.21354179484688512, + "learning_rate": 3.5120731249694114e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40277063846588135, + "step": 5640, + "valid_targets_mean": 28776.8, + "valid_targets_min": 19899 + }, + { + "epoch": 6.024546424759872, + "grad_norm": 0.21073033421592155, + "learning_rate": 3.494371764602029e-05, + "loss": 0.4038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4004727602005005, + "step": 5645, + "valid_targets_mean": 28738.6, + "valid_targets_min": 21317 + }, + { + "epoch": 6.029882604055496, + "grad_norm": 0.27975450593983187, + "learning_rate": 3.476705676725218e-05, + "loss": 0.4029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4047626852989197, + "step": 5650, + "valid_targets_mean": 28777.9, + "valid_targets_min": 22491 + }, + { + "epoch": 6.03521878335112, + "grad_norm": 0.2253042599562701, + "learning_rate": 3.459074957121713e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40415048599243164, + "step": 5655, + "valid_targets_mean": 28801.4, + "valid_targets_min": 20094 + }, + { + "epoch": 6.040554962646745, + "grad_norm": 0.23375326699273516, + "learning_rate": 3.441479701382504e-05, + "loss": 0.4011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39832621812820435, + "step": 5660, + "valid_targets_mean": 28750.8, + "valid_targets_min": 19771 + }, + { + "epoch": 6.045891141942369, + "grad_norm": 0.2198729246111239, + "learning_rate": 3.423920004906289e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40202754735946655, + "step": 5665, + "valid_targets_mean": 28771.3, + "valid_targets_min": 20397 + }, + { + "epoch": 6.0512273212379935, + "grad_norm": 0.22331278226059176, + "learning_rate": 3.4063959628989684e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40373921394348145, + "step": 5670, + "valid_targets_mean": 28709.1, + "valid_targets_min": 19664 + }, + { + "epoch": 6.056563500533618, + "grad_norm": 0.2094224833863894, + "learning_rate": 3.388907670373147e-05, + "loss": 0.4038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40180504322052, + "step": 5675, + "valid_targets_mean": 28815.6, + "valid_targets_min": 20582 + }, + { + "epoch": 6.061899679829242, + "grad_norm": 0.25422996974709433, + "learning_rate": 3.3714552221475805e-05, + "loss": 0.4035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4039395749568939, + "step": 5680, + "valid_targets_mean": 28961.4, + "valid_targets_min": 21772 + }, + { + "epoch": 6.0672358591248665, + "grad_norm": 0.19409749836193413, + "learning_rate": 3.354038712846698e-05, + "loss": 0.4039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40708667039871216, + "step": 5685, + "valid_targets_mean": 28967.0, + "valid_targets_min": 19766 + }, + { + "epoch": 6.072572038420491, + "grad_norm": 0.23052130057371095, + "learning_rate": 3.336658236900062e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4013357162475586, + "step": 5690, + "valid_targets_mean": 28754.0, + "valid_targets_min": 21630 + }, + { + "epoch": 6.077908217716115, + "grad_norm": 0.2231430978668057, + "learning_rate": 3.3193138885418715e-05, + "loss": 0.4035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4056231379508972, + "step": 5695, + "valid_targets_mean": 28825.4, + "valid_targets_min": 21956 + }, + { + "epoch": 6.08324439701174, + "grad_norm": 0.21788060543120694, + "learning_rate": 3.302005761810453e-05, + "loss": 0.4045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4046787917613983, + "step": 5700, + "valid_targets_mean": 28996.1, + "valid_targets_min": 19864 + }, + { + "epoch": 6.088580576307364, + "grad_norm": 0.22295642951669387, + "learning_rate": 3.2847339505477374e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4034949541091919, + "step": 5705, + "valid_targets_mean": 28886.7, + "valid_targets_min": 18510 + }, + { + "epoch": 6.093916755602988, + "grad_norm": 0.23010510678057203, + "learning_rate": 3.267498548398766e-05, + "loss": 0.4052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40143734216690063, + "step": 5710, + "valid_targets_mean": 28961.2, + "valid_targets_min": 19954 + }, + { + "epoch": 6.099252934898613, + "grad_norm": 0.20873053646578818, + "learning_rate": 3.250299648811168e-05, + "loss": 0.4023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40607360005378723, + "step": 5715, + "valid_targets_mean": 28922.8, + "valid_targets_min": 19932 + }, + { + "epoch": 6.104589114194237, + "grad_norm": 0.2083861463262958, + "learning_rate": 3.233137345034669e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40308186411857605, + "step": 5720, + "valid_targets_mean": 28935.9, + "valid_targets_min": 21625 + }, + { + "epoch": 6.109925293489861, + "grad_norm": 0.2043950913043271, + "learning_rate": 3.2160117301205726e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40011361241340637, + "step": 5725, + "valid_targets_mean": 28786.4, + "valid_targets_min": 16026 + }, + { + "epoch": 6.115261472785486, + "grad_norm": 0.20300023613652335, + "learning_rate": 3.198922896921259e-05, + "loss": 0.4023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40243393182754517, + "step": 5730, + "valid_targets_mean": 28795.3, + "valid_targets_min": 21856 + }, + { + "epoch": 6.12059765208111, + "grad_norm": 0.20243128119834547, + "learning_rate": 3.1818709380896905e-05, + "loss": 0.4037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3996739983558655, + "step": 5735, + "valid_targets_mean": 28954.9, + "valid_targets_min": 20610 + }, + { + "epoch": 6.125933831376734, + "grad_norm": 0.20548029321838865, + "learning_rate": 3.164855946078899e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40435925126075745, + "step": 5740, + "valid_targets_mean": 28726.2, + "valid_targets_min": 16033 + }, + { + "epoch": 6.131270010672359, + "grad_norm": 0.20211522984866082, + "learning_rate": 3.147878013141492e-05, + "loss": 0.4004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39486247301101685, + "step": 5745, + "valid_targets_mean": 29027.4, + "valid_targets_min": 23070 + }, + { + "epoch": 6.136606189967983, + "grad_norm": 0.2127912730791383, + "learning_rate": 3.130937231329142e-05, + "loss": 0.4037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40694287419319153, + "step": 5750, + "valid_targets_mean": 28807.9, + "valid_targets_min": 20486 + }, + { + "epoch": 6.1419423692636075, + "grad_norm": 0.20031145012395565, + "learning_rate": 3.114033692492093e-05, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4004095196723938, + "step": 5755, + "valid_targets_mean": 28869.6, + "valid_targets_min": 20911 + }, + { + "epoch": 6.147278548559232, + "grad_norm": 0.20549511785676874, + "learning_rate": 3.097167488278672e-05, + "loss": 0.4024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4021293520927429, + "step": 5760, + "valid_targets_mean": 28929.9, + "valid_targets_min": 20613 + }, + { + "epoch": 6.152614727854856, + "grad_norm": 0.20927356784178391, + "learning_rate": 3.080338710134778e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40633559226989746, + "step": 5765, + "valid_targets_mean": 29096.6, + "valid_targets_min": 21570 + }, + { + "epoch": 6.1579509071504805, + "grad_norm": 0.21312244449480242, + "learning_rate": 3.063547449303397e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40091001987457275, + "step": 5770, + "valid_targets_mean": 28731.6, + "valid_targets_min": 21337 + }, + { + "epoch": 6.163287086446105, + "grad_norm": 0.21230355994256525, + "learning_rate": 3.04679379682409e-05, + "loss": 0.402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40138083696365356, + "step": 5775, + "valid_targets_mean": 28948.2, + "valid_targets_min": 23092 + }, + { + "epoch": 6.168623265741729, + "grad_norm": 0.1976015266522106, + "learning_rate": 3.030077843532526e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40313124656677246, + "step": 5780, + "valid_targets_mean": 28983.9, + "valid_targets_min": 20284 + }, + { + "epoch": 6.173959445037354, + "grad_norm": 0.20487890681336945, + "learning_rate": 3.013399680059965e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4016597867012024, + "step": 5785, + "valid_targets_mean": 28889.0, + "valid_targets_min": 18214 + }, + { + "epoch": 6.179295624332978, + "grad_norm": 0.19989947369680958, + "learning_rate": 2.996759396832778e-05, + "loss": 0.4011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40249523520469666, + "step": 5790, + "valid_targets_mean": 28888.5, + "valid_targets_min": 22390 + }, + { + "epoch": 6.184631803628602, + "grad_norm": 0.19919529330401847, + "learning_rate": 2.980157084071957e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4005407691001892, + "step": 5795, + "valid_targets_mean": 28789.4, + "valid_targets_min": 20334 + }, + { + "epoch": 6.189967982924227, + "grad_norm": 0.2221574146814808, + "learning_rate": 2.9635928317926277e-05, + "loss": 0.4007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40457186102867126, + "step": 5800, + "valid_targets_mean": 28785.1, + "valid_targets_min": 17119 + }, + { + "epoch": 6.19530416221985, + "grad_norm": 0.1990827141497274, + "learning_rate": 2.9470667298035558e-05, + "loss": 0.4054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40230417251586914, + "step": 5805, + "valid_targets_mean": 28801.1, + "valid_targets_min": 20516 + }, + { + "epoch": 6.2006403415154745, + "grad_norm": 0.21931541168701757, + "learning_rate": 2.930578867706657e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39906126260757446, + "step": 5810, + "valid_targets_mean": 28834.8, + "valid_targets_min": 20373 + }, + { + "epoch": 6.205976520811099, + "grad_norm": 0.19070110342281413, + "learning_rate": 2.9141293348965183e-05, + "loss": 0.4053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4032185971736908, + "step": 5815, + "valid_targets_mean": 28856.1, + "valid_targets_min": 19452 + }, + { + "epoch": 6.211312700106723, + "grad_norm": 0.2070312687844574, + "learning_rate": 2.8977182205599184e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40679287910461426, + "step": 5820, + "valid_targets_mean": 28810.0, + "valid_targets_min": 20350 + }, + { + "epoch": 6.216648879402348, + "grad_norm": 0.1973782077994425, + "learning_rate": 2.8813456136753213e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4033341407775879, + "step": 5825, + "valid_targets_mean": 28874.5, + "valid_targets_min": 18152 + }, + { + "epoch": 6.221985058697972, + "grad_norm": 0.2137726492190117, + "learning_rate": 2.8650116030124353e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40437471866607666, + "step": 5830, + "valid_targets_mean": 28704.6, + "valid_targets_min": 22485 + }, + { + "epoch": 6.227321237993596, + "grad_norm": 0.2234305179964095, + "learning_rate": 2.8487162771316787e-05, + "loss": 0.4012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40185803174972534, + "step": 5835, + "valid_targets_mean": 28684.6, + "valid_targets_min": 21339 + }, + { + "epoch": 6.232657417289221, + "grad_norm": 0.18891512696399268, + "learning_rate": 2.832459724383748e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.400049090385437, + "step": 5840, + "valid_targets_mean": 28835.2, + "valid_targets_min": 20750 + }, + { + "epoch": 6.237993596584845, + "grad_norm": 0.1941215255944952, + "learning_rate": 2.8162420329091034e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4041074812412262, + "step": 5845, + "valid_targets_mean": 28848.6, + "valid_targets_min": 21970 + }, + { + "epoch": 6.243329775880469, + "grad_norm": 0.20108340167380698, + "learning_rate": 2.80006329063751e-05, + "loss": 0.404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40438374876976013, + "step": 5850, + "valid_targets_mean": 28858.2, + "valid_targets_min": 22768 + }, + { + "epoch": 6.248665955176094, + "grad_norm": 0.18986968645064523, + "learning_rate": 2.783923585287559e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40219926834106445, + "step": 5855, + "valid_targets_mean": 28752.1, + "valid_targets_min": 22450 + }, + { + "epoch": 6.254002134471718, + "grad_norm": 0.19131675041327859, + "learning_rate": 2.76782300436619e-05, + "loss": 0.4032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40058252215385437, + "step": 5860, + "valid_targets_mean": 28910.0, + "valid_targets_min": 23690 + }, + { + "epoch": 6.259338313767342, + "grad_norm": 0.20587972668628907, + "learning_rate": 2.7517616351682153e-05, + "loss": 0.4038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40199992060661316, + "step": 5865, + "valid_targets_mean": 28825.8, + "valid_targets_min": 21820 + }, + { + "epoch": 6.264674493062967, + "grad_norm": 0.22498247411615535, + "learning_rate": 2.7357395647758454e-05, + "loss": 0.4044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4067252576351166, + "step": 5870, + "valid_targets_mean": 28796.7, + "valid_targets_min": 21420 + }, + { + "epoch": 6.270010672358591, + "grad_norm": 0.18859149173847364, + "learning_rate": 2.719756880058214e-05, + "loss": 0.4035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4027019441127777, + "step": 5875, + "valid_targets_mean": 28784.4, + "valid_targets_min": 18739 + }, + { + "epoch": 6.2753468516542155, + "grad_norm": 0.21404109929128848, + "learning_rate": 2.7038136676709268e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40717530250549316, + "step": 5880, + "valid_targets_mean": 28924.2, + "valid_targets_min": 18971 + }, + { + "epoch": 6.28068303094984, + "grad_norm": 0.2130469328147773, + "learning_rate": 2.6879100140555592e-05, + "loss": 0.4032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4079652428627014, + "step": 5885, + "valid_targets_mean": 28976.3, + "valid_targets_min": 22673 + }, + { + "epoch": 6.286019210245464, + "grad_norm": 0.2008718404415538, + "learning_rate": 2.6720460054392183e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40473008155822754, + "step": 5890, + "valid_targets_mean": 28953.7, + "valid_targets_min": 20946 + }, + { + "epoch": 6.2913553895410885, + "grad_norm": 0.21367693351351572, + "learning_rate": 2.656221727834056e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4048454165458679, + "step": 5895, + "valid_targets_mean": 28883.1, + "valid_targets_min": 22808 + }, + { + "epoch": 6.296691568836713, + "grad_norm": 0.18360734367399942, + "learning_rate": 2.640437267036815e-05, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40429389476776123, + "step": 5900, + "valid_targets_mean": 28847.2, + "valid_targets_min": 21654 + }, + { + "epoch": 6.302027748132337, + "grad_norm": 0.19376965246444747, + "learning_rate": 2.624692708628349e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40079379081726074, + "step": 5905, + "valid_targets_mean": 28974.9, + "valid_targets_min": 21683 + }, + { + "epoch": 6.307363927427962, + "grad_norm": 0.1980799701228368, + "learning_rate": 2.6089881379731684e-05, + "loss": 0.4037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40692567825317383, + "step": 5910, + "valid_targets_mean": 28900.9, + "valid_targets_min": 21402 + }, + { + "epoch": 6.312700106723586, + "grad_norm": 0.1805305637412513, + "learning_rate": 2.593323640218983e-05, + "loss": 0.4032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.403741717338562, + "step": 5915, + "valid_targets_mean": 28839.7, + "valid_targets_min": 22345 + }, + { + "epoch": 6.31803628601921, + "grad_norm": 0.17499517264808107, + "learning_rate": 2.577699300296229e-05, + "loss": 0.402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4021400511264801, + "step": 5920, + "valid_targets_mean": 28891.9, + "valid_targets_min": 21526 + }, + { + "epoch": 6.323372465314835, + "grad_norm": 0.17937660832698205, + "learning_rate": 2.5621152029176175e-05, + "loss": 0.4013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4012629985809326, + "step": 5925, + "valid_targets_mean": 28952.7, + "valid_targets_min": 22270 + }, + { + "epoch": 6.328708644610459, + "grad_norm": 0.18444591108122746, + "learning_rate": 2.5465714325776645e-05, + "loss": 0.4028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40533414483070374, + "step": 5930, + "valid_targets_mean": 28918.4, + "valid_targets_min": 21858 + }, + { + "epoch": 6.334044823906083, + "grad_norm": 0.18379872762780242, + "learning_rate": 2.5310680735522385e-05, + "loss": 0.402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40175458788871765, + "step": 5935, + "valid_targets_mean": 28804.4, + "valid_targets_min": 19394 + }, + { + "epoch": 6.339381003201708, + "grad_norm": 0.18930997642920325, + "learning_rate": 2.5156052098981143e-05, + "loss": 0.4038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40313613414764404, + "step": 5940, + "valid_targets_mean": 28761.6, + "valid_targets_min": 19766 + }, + { + "epoch": 6.344717182497332, + "grad_norm": 0.1832631480960261, + "learning_rate": 2.5001829254524945e-05, + "loss": 0.4032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40204566717147827, + "step": 5945, + "valid_targets_mean": 28999.8, + "valid_targets_min": 21177 + }, + { + "epoch": 6.350053361792956, + "grad_norm": 0.2273332984394366, + "learning_rate": 2.4848013038325756e-05, + "loss": 0.4039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40472328662872314, + "step": 5950, + "valid_targets_mean": 28942.7, + "valid_targets_min": 21101 + }, + { + "epoch": 6.355389541088581, + "grad_norm": 0.19163706801171704, + "learning_rate": 2.4694604284350875e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39981400966644287, + "step": 5955, + "valid_targets_mean": 28912.2, + "valid_targets_min": 18845 + }, + { + "epoch": 6.360725720384205, + "grad_norm": 0.19069414768695025, + "learning_rate": 2.4541603824358384e-05, + "loss": 0.4022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.402090847492218, + "step": 5960, + "valid_targets_mean": 28865.6, + "valid_targets_min": 23438 + }, + { + "epoch": 6.3660618996798295, + "grad_norm": 0.18179804752403267, + "learning_rate": 2.4389012487892627e-05, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40023428201675415, + "step": 5965, + "valid_targets_mean": 28759.3, + "valid_targets_min": 21201 + }, + { + "epoch": 6.371398078975454, + "grad_norm": 0.2148975593144794, + "learning_rate": 2.4236831102279756e-05, + "loss": 0.4022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40240776538848877, + "step": 5970, + "valid_targets_mean": 28966.0, + "valid_targets_min": 22745 + }, + { + "epoch": 6.376734258271078, + "grad_norm": 0.18490174473146967, + "learning_rate": 2.408506049262329e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4022800028324127, + "step": 5975, + "valid_targets_mean": 28688.3, + "valid_targets_min": 21559 + }, + { + "epoch": 6.382070437566703, + "grad_norm": 0.1844255041375129, + "learning_rate": 2.3933701481799496e-05, + "loss": 0.401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40320301055908203, + "step": 5980, + "valid_targets_mean": 28579.6, + "valid_targets_min": 20348 + }, + { + "epoch": 6.387406616862327, + "grad_norm": 0.21357911835637616, + "learning_rate": 2.3782754890453106e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3996123671531677, + "step": 5985, + "valid_targets_mean": 28688.8, + "valid_targets_min": 19354 + }, + { + "epoch": 6.392742796157951, + "grad_norm": 0.196175539187502, + "learning_rate": 2.363222153699276e-05, + "loss": 0.405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4055750072002411, + "step": 5990, + "valid_targets_mean": 28719.9, + "valid_targets_min": 20713 + }, + { + "epoch": 6.398078975453576, + "grad_norm": 0.18380230078382345, + "learning_rate": 2.34821022375865e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4092828333377838, + "step": 5995, + "valid_targets_mean": 28915.8, + "valid_targets_min": 23482 + }, + { + "epoch": 6.4034151547492, + "grad_norm": 0.16559855043636187, + "learning_rate": 2.333239780615756e-05, + "loss": 0.4023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40028464794158936, + "step": 6000, + "valid_targets_mean": 28787.5, + "valid_targets_min": 21447 + }, + { + "epoch": 6.408751334044824, + "grad_norm": 0.19131444987514326, + "learning_rate": 2.3183109054379715e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40193384885787964, + "step": 6005, + "valid_targets_mean": 28890.1, + "valid_targets_min": 21725 + }, + { + "epoch": 6.414087513340448, + "grad_norm": 0.20081963856111104, + "learning_rate": 2.3034236791673058e-05, + "loss": 0.4012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40240931510925293, + "step": 6010, + "valid_targets_mean": 28768.8, + "valid_targets_min": 21470 + }, + { + "epoch": 6.419423692636073, + "grad_norm": 0.19922149504539194, + "learning_rate": 2.2885781825199526e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4008885622024536, + "step": 6015, + "valid_targets_mean": 28945.8, + "valid_targets_min": 22275 + }, + { + "epoch": 6.4247598719316965, + "grad_norm": 0.19131542925233516, + "learning_rate": 2.2737744959858577e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3999541401863098, + "step": 6020, + "valid_targets_mean": 28794.4, + "valid_targets_min": 21997 + }, + { + "epoch": 6.430096051227321, + "grad_norm": 0.19508531274083823, + "learning_rate": 2.2590126998282725e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4026764929294586, + "step": 6025, + "valid_targets_mean": 28815.1, + "valid_targets_min": 20115 + }, + { + "epoch": 6.435432230522945, + "grad_norm": 0.1817251897940026, + "learning_rate": 2.2442928740833245e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4037598669528961, + "step": 6030, + "valid_targets_mean": 28978.1, + "valid_targets_min": 22243 + }, + { + "epoch": 6.44076840981857, + "grad_norm": 0.1877156586444744, + "learning_rate": 2.229615098559594e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4073331356048584, + "step": 6035, + "valid_targets_mean": 28778.0, + "valid_targets_min": 21732 + }, + { + "epoch": 6.446104589114194, + "grad_norm": 0.1883960030591332, + "learning_rate": 2.214979452837661e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39945822954177856, + "step": 6040, + "valid_targets_mean": 28708.4, + "valid_targets_min": 22806 + }, + { + "epoch": 6.451440768409818, + "grad_norm": 0.18792510830856934, + "learning_rate": 2.200386016269691e-05, + "loss": 0.4027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40260523557662964, + "step": 6045, + "valid_targets_mean": 28852.8, + "valid_targets_min": 23161 + }, + { + "epoch": 6.456776947705443, + "grad_norm": 0.18046002723349452, + "learning_rate": 2.1858348679789987e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40115612745285034, + "step": 6050, + "valid_targets_mean": 28744.5, + "valid_targets_min": 17458 + }, + { + "epoch": 6.462113127001067, + "grad_norm": 0.19326827523187853, + "learning_rate": 2.171326086859612e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4018067717552185, + "step": 6055, + "valid_targets_mean": 28825.2, + "valid_targets_min": 20330 + }, + { + "epoch": 6.467449306296691, + "grad_norm": 0.18187928385392707, + "learning_rate": 2.1568597515758572e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3992971181869507, + "step": 6060, + "valid_targets_mean": 28750.9, + "valid_targets_min": 22831 + }, + { + "epoch": 6.472785485592316, + "grad_norm": 0.18571117091706152, + "learning_rate": 2.142435940561921e-05, + "loss": 0.4026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40224558115005493, + "step": 6065, + "valid_targets_mean": 28840.0, + "valid_targets_min": 23005 + }, + { + "epoch": 6.47812166488794, + "grad_norm": 0.19856471998068015, + "learning_rate": 2.128054732021435e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4016802906990051, + "step": 6070, + "valid_targets_mean": 28950.6, + "valid_targets_min": 21925 + }, + { + "epoch": 6.483457844183564, + "grad_norm": 0.18011864362954913, + "learning_rate": 2.1137162039270386e-05, + "loss": 0.4032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40601515769958496, + "step": 6075, + "valid_targets_mean": 28755.4, + "valid_targets_min": 20836 + }, + { + "epoch": 6.488794023479189, + "grad_norm": 0.1754307871285436, + "learning_rate": 2.0994204340199806e-05, + "loss": 0.402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40100088715553284, + "step": 6080, + "valid_targets_mean": 28894.5, + "valid_targets_min": 21107 + }, + { + "epoch": 6.494130202774813, + "grad_norm": 0.18780028341035576, + "learning_rate": 2.085167499809666e-05, + "loss": 0.4009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40386563539505005, + "step": 6085, + "valid_targets_mean": 28856.9, + "valid_targets_min": 20525 + }, + { + "epoch": 6.4994663820704375, + "grad_norm": 0.17939591035162056, + "learning_rate": 2.0709574785732544e-05, + "loss": 0.4043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40500178933143616, + "step": 6090, + "valid_targets_mean": 28813.9, + "valid_targets_min": 22234 + }, + { + "epoch": 6.504802561366062, + "grad_norm": 0.19077799045157887, + "learning_rate": 2.0567904473552446e-05, + "loss": 0.4023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4002484083175659, + "step": 6095, + "valid_targets_mean": 28850.9, + "valid_targets_min": 21860 + }, + { + "epoch": 6.510138740661686, + "grad_norm": 0.1897438513167475, + "learning_rate": 2.0426664829670415e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.399167537689209, + "step": 6100, + "valid_targets_mean": 29116.8, + "valid_targets_min": 23393 + }, + { + "epoch": 6.5154749199573105, + "grad_norm": 0.19359987172136103, + "learning_rate": 2.0285856619865573e-05, + "loss": 0.3993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4001930356025696, + "step": 6105, + "valid_targets_mean": 28885.3, + "valid_targets_min": 22689 + }, + { + "epoch": 6.520811099252935, + "grad_norm": 0.1837096087171172, + "learning_rate": 2.014548060757785e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4038437306880951, + "step": 6110, + "valid_targets_mean": 28748.8, + "valid_targets_min": 17902 + }, + { + "epoch": 6.526147278548559, + "grad_norm": 0.17583615950748377, + "learning_rate": 2.0005537553903787e-05, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40059417486190796, + "step": 6115, + "valid_targets_mean": 28806.0, + "valid_targets_min": 17871 + }, + { + "epoch": 6.531483457844184, + "grad_norm": 0.1817073021255667, + "learning_rate": 1.9866028217592612e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3994242548942566, + "step": 6120, + "valid_targets_mean": 28856.7, + "valid_targets_min": 19786 + }, + { + "epoch": 6.536819637139808, + "grad_norm": 0.18507873365948924, + "learning_rate": 1.972695335504192e-05, + "loss": 0.4025, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3999413251876831, + "step": 6125, + "valid_targets_mean": 28968.7, + "valid_targets_min": 23619 + }, + { + "epoch": 6.542155816435432, + "grad_norm": 0.17834966809846056, + "learning_rate": 1.9588313720293717e-05, + "loss": 0.4029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40515387058258057, + "step": 6130, + "valid_targets_mean": 28770.1, + "valid_targets_min": 19698 + }, + { + "epoch": 6.547491995731057, + "grad_norm": 0.17179000860676627, + "learning_rate": 1.9450110065030202e-05, + "loss": 0.4008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3992699980735779, + "step": 6135, + "valid_targets_mean": 28938.3, + "valid_targets_min": 16427 + }, + { + "epoch": 6.552828175026681, + "grad_norm": 0.18437786009110965, + "learning_rate": 1.9312343138569822e-05, + "loss": 0.4024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.400585412979126, + "step": 6140, + "valid_targets_mean": 28727.0, + "valid_targets_min": 20169 + }, + { + "epoch": 6.558164354322305, + "grad_norm": 0.19252564397052424, + "learning_rate": 1.9175013687863165e-05, + "loss": 0.4007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40095585584640503, + "step": 6145, + "valid_targets_mean": 28912.5, + "valid_targets_min": 21936 + }, + { + "epoch": 6.56350053361793, + "grad_norm": 0.1816462272421975, + "learning_rate": 1.9038122457488804e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39897194504737854, + "step": 6150, + "valid_targets_mean": 28958.5, + "valid_targets_min": 22794 + }, + { + "epoch": 6.568836712913554, + "grad_norm": 0.18775048667792404, + "learning_rate": 1.8901670189649457e-05, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40145641565322876, + "step": 6155, + "valid_targets_mean": 29103.2, + "valid_targets_min": 22797 + }, + { + "epoch": 6.574172892209178, + "grad_norm": 0.19158581227718158, + "learning_rate": 1.8765657624167765e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40386325120925903, + "step": 6160, + "valid_targets_mean": 28646.2, + "valid_targets_min": 22585 + }, + { + "epoch": 6.579509071504803, + "grad_norm": 0.18156089305735862, + "learning_rate": 1.863008549848243e-05, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4037448763847351, + "step": 6165, + "valid_targets_mean": 28938.0, + "valid_targets_min": 21285 + }, + { + "epoch": 6.584845250800427, + "grad_norm": 0.19574581682266376, + "learning_rate": 1.849495454764416e-05, + "loss": 0.4013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39787396788597107, + "step": 6170, + "valid_targets_mean": 28945.8, + "valid_targets_min": 19712 + }, + { + "epoch": 6.5901814300960515, + "grad_norm": 0.18243929360702177, + "learning_rate": 1.8360265504311635e-05, + "loss": 0.402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40304043889045715, + "step": 6175, + "valid_targets_mean": 28608.3, + "valid_targets_min": 20646 + }, + { + "epoch": 6.595517609391676, + "grad_norm": 0.18102304076902928, + "learning_rate": 1.822601909874765e-05, + "loss": 0.4025, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40229958295822144, + "step": 6180, + "valid_targets_mean": 28912.8, + "valid_targets_min": 22071 + }, + { + "epoch": 6.6008537886873, + "grad_norm": 0.18009646165533064, + "learning_rate": 1.8092216058814993e-05, + "loss": 0.4013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4023338556289673, + "step": 6185, + "valid_targets_mean": 28818.4, + "valid_targets_min": 19823 + }, + { + "epoch": 6.606189967982925, + "grad_norm": 0.17824539351505556, + "learning_rate": 1.7958857109972714e-05, + "loss": 0.4007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39848190546035767, + "step": 6190, + "valid_targets_mean": 28675.4, + "valid_targets_min": 21687 + }, + { + "epoch": 6.611526147278549, + "grad_norm": 0.17141496286854313, + "learning_rate": 1.7825942975271946e-05, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39964133501052856, + "step": 6195, + "valid_targets_mean": 28842.7, + "valid_targets_min": 22278 + }, + { + "epoch": 6.616862326574173, + "grad_norm": 0.18120109301172674, + "learning_rate": 1.7693474375352193e-05, + "loss": 0.4014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40147751569747925, + "step": 6200, + "valid_targets_mean": 28814.5, + "valid_targets_min": 22127 + }, + { + "epoch": 6.622198505869797, + "grad_norm": 0.18343171990938137, + "learning_rate": 1.7561452028437362e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40402135252952576, + "step": 6205, + "valid_targets_mean": 28789.7, + "valid_targets_min": 19184 + }, + { + "epoch": 6.627534685165422, + "grad_norm": 0.1617202995515404, + "learning_rate": 1.7429876650331732e-05, + "loss": 0.402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.399649441242218, + "step": 6210, + "valid_targets_mean": 28866.6, + "valid_targets_min": 23176 + }, + { + "epoch": 6.6328708644610455, + "grad_norm": 0.18035454442476903, + "learning_rate": 1.7298748954416334e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.406667023897171, + "step": 6215, + "valid_targets_mean": 28882.5, + "valid_targets_min": 22635 + }, + { + "epoch": 6.638207043756671, + "grad_norm": 0.16518628892225848, + "learning_rate": 1.716806965164477e-05, + "loss": 0.4004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4044821858406067, + "step": 6220, + "valid_targets_mean": 28862.5, + "valid_targets_min": 20417 + }, + { + "epoch": 6.643543223052294, + "grad_norm": 0.17373762749649188, + "learning_rate": 1.7037839450539706e-05, + "loss": 0.3983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39518213272094727, + "step": 6225, + "valid_targets_mean": 28887.1, + "valid_targets_min": 16333 + }, + { + "epoch": 6.6488794023479185, + "grad_norm": 0.1766409075861455, + "learning_rate": 1.6908059057188686e-05, + "loss": 0.4032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40541961789131165, + "step": 6230, + "valid_targets_mean": 28740.7, + "valid_targets_min": 19735 + }, + { + "epoch": 6.654215581643543, + "grad_norm": 0.1968366034578871, + "learning_rate": 1.677872917524057e-05, + "loss": 0.3994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4014188051223755, + "step": 6235, + "valid_targets_mean": 28856.3, + "valid_targets_min": 20610 + }, + { + "epoch": 6.659551760939167, + "grad_norm": 0.18237048405103176, + "learning_rate": 1.664985050590162e-05, + "loss": 0.4022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40323466062545776, + "step": 6240, + "valid_targets_mean": 28779.6, + "valid_targets_min": 20577 + }, + { + "epoch": 6.664887940234792, + "grad_norm": 0.2157169595866924, + "learning_rate": 1.6521423747931597e-05, + "loss": 0.4026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39827534556388855, + "step": 6245, + "valid_targets_mean": 28740.5, + "valid_targets_min": 21039 + }, + { + "epoch": 6.670224119530416, + "grad_norm": 0.16378620466568197, + "learning_rate": 1.63934495976402e-05, + "loss": 0.4024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40637755393981934, + "step": 6250, + "valid_targets_mean": 28883.8, + "valid_targets_min": 20594 + }, + { + "epoch": 6.67556029882604, + "grad_norm": 0.19308134301794055, + "learning_rate": 1.626592874888303e-05, + "loss": 0.4036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4051148295402527, + "step": 6255, + "valid_targets_mean": 28807.3, + "valid_targets_min": 22866 + }, + { + "epoch": 6.680896478121665, + "grad_norm": 0.15946375328063236, + "learning_rate": 1.6138861893058067e-05, + "loss": 0.4029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4002209007740021, + "step": 6260, + "valid_targets_mean": 28840.7, + "valid_targets_min": 17736 + }, + { + "epoch": 6.686232657417289, + "grad_norm": 0.1631580545072671, + "learning_rate": 1.601224971910177e-05, + "loss": 0.4023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40180450677871704, + "step": 6265, + "valid_targets_mean": 28697.5, + "valid_targets_min": 17344 + }, + { + "epoch": 6.691568836712913, + "grad_norm": 0.18043079602468606, + "learning_rate": 1.588609291348535e-05, + "loss": 0.401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.401018887758255, + "step": 6270, + "valid_targets_mean": 28887.6, + "valid_targets_min": 20293 + }, + { + "epoch": 6.696905016008538, + "grad_norm": 0.19874888578672376, + "learning_rate": 1.5760392160211156e-05, + "loss": 0.4013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3999865651130676, + "step": 6275, + "valid_targets_mean": 28792.9, + "valid_targets_min": 21576 + }, + { + "epoch": 6.702241195304162, + "grad_norm": 0.17033556307489203, + "learning_rate": 1.563514814080881e-05, + "loss": 0.4004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4025978744029999, + "step": 6280, + "valid_targets_mean": 28552.2, + "valid_targets_min": 21530 + }, + { + "epoch": 6.707577374599786, + "grad_norm": 0.18220728792481394, + "learning_rate": 1.5510361534331686e-05, + "loss": 0.4022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4051911234855652, + "step": 6285, + "valid_targets_mean": 28988.1, + "valid_targets_min": 19604 + }, + { + "epoch": 6.712913553895411, + "grad_norm": 0.1979186351056776, + "learning_rate": 1.538603301735305e-05, + "loss": 0.4006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3975885510444641, + "step": 6290, + "valid_targets_mean": 29006.6, + "valid_targets_min": 20821 + }, + { + "epoch": 6.718249733191035, + "grad_norm": 0.18776718169818407, + "learning_rate": 1.5262163263962527e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4064946174621582, + "step": 6295, + "valid_targets_mean": 28904.0, + "valid_targets_min": 22811 + }, + { + "epoch": 6.7235859124866595, + "grad_norm": 0.18165907356672129, + "learning_rate": 1.5138752945762425e-05, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40303996205329895, + "step": 6300, + "valid_targets_mean": 28870.6, + "valid_targets_min": 20970 + }, + { + "epoch": 6.728922091782284, + "grad_norm": 0.17790068533186368, + "learning_rate": 1.5015802731863981e-05, + "loss": 0.3988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39912688732147217, + "step": 6305, + "valid_targets_mean": 28779.1, + "valid_targets_min": 20784 + }, + { + "epoch": 6.734258271077908, + "grad_norm": 0.16495901813636496, + "learning_rate": 1.4893313288883915e-05, + "loss": 0.401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4006229043006897, + "step": 6310, + "valid_targets_mean": 28860.0, + "valid_targets_min": 18795 + }, + { + "epoch": 6.7395944503735326, + "grad_norm": 0.16451303664483405, + "learning_rate": 1.4771285280940616e-05, + "loss": 0.4005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40138137340545654, + "step": 6315, + "valid_targets_mean": 28857.6, + "valid_targets_min": 19825 + }, + { + "epoch": 6.744930629669157, + "grad_norm": 0.17086287887312568, + "learning_rate": 1.4649719369650772e-05, + "loss": 0.3999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40100106596946716, + "step": 6320, + "valid_targets_mean": 28765.1, + "valid_targets_min": 18374 + }, + { + "epoch": 6.750266808964781, + "grad_norm": 0.16714376991812513, + "learning_rate": 1.4528616214125535e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4033486247062683, + "step": 6325, + "valid_targets_mean": 28851.5, + "valid_targets_min": 21665 + }, + { + "epoch": 6.755602988260406, + "grad_norm": 0.16704078585302812, + "learning_rate": 1.4407976470967133e-05, + "loss": 0.4025, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40259772539138794, + "step": 6330, + "valid_targets_mean": 28977.0, + "valid_targets_min": 21840 + }, + { + "epoch": 6.76093916755603, + "grad_norm": 0.16866547396425438, + "learning_rate": 1.428780079426526e-05, + "loss": 0.4026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4046093225479126, + "step": 6335, + "valid_targets_mean": 28653.1, + "valid_targets_min": 18063 + }, + { + "epoch": 6.766275346851654, + "grad_norm": 0.16086409262800527, + "learning_rate": 1.4168089835593446e-05, + "loss": 0.4018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40628403425216675, + "step": 6340, + "valid_targets_mean": 28960.2, + "valid_targets_min": 20183 + }, + { + "epoch": 6.771611526147279, + "grad_norm": 0.1869500620775073, + "learning_rate": 1.4048844244005666e-05, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4005264341831207, + "step": 6345, + "valid_targets_mean": 28785.8, + "valid_targets_min": 23578 + }, + { + "epoch": 6.776947705442903, + "grad_norm": 0.1697742455933057, + "learning_rate": 1.3930064666032693e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4072091579437256, + "step": 6350, + "valid_targets_mean": 28691.9, + "valid_targets_min": 22316 + }, + { + "epoch": 6.782283884738527, + "grad_norm": 0.16988728925311614, + "learning_rate": 1.381175174567868e-05, + "loss": 0.4037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4012793302536011, + "step": 6355, + "valid_targets_mean": 28903.2, + "valid_targets_min": 22100 + }, + { + "epoch": 6.787620064034152, + "grad_norm": 0.1642169911188869, + "learning_rate": 1.3693906124417655e-05, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4029124975204468, + "step": 6360, + "valid_targets_mean": 28804.8, + "valid_targets_min": 22231 + }, + { + "epoch": 6.792956243329776, + "grad_norm": 0.18018467005146654, + "learning_rate": 1.357652844118994e-05, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40579214692115784, + "step": 6365, + "valid_targets_mean": 28713.4, + "valid_targets_min": 19188 + }, + { + "epoch": 6.7982924226254005, + "grad_norm": 0.18591848673587488, + "learning_rate": 1.3459619332398887e-05, + "loss": 0.4025, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4033072888851166, + "step": 6370, + "valid_targets_mean": 28754.1, + "valid_targets_min": 20035 + }, + { + "epoch": 6.803628601921025, + "grad_norm": 0.14922686750987793, + "learning_rate": 1.3343179431907193e-05, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3995817303657532, + "step": 6375, + "valid_targets_mean": 28785.5, + "valid_targets_min": 21644 + }, + { + "epoch": 6.808964781216649, + "grad_norm": 0.16633454388559174, + "learning_rate": 1.32272093710337e-05, + "loss": 0.4, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3998081088066101, + "step": 6380, + "valid_targets_mean": 28761.7, + "valid_targets_min": 21351 + }, + { + "epoch": 6.8143009605122735, + "grad_norm": 0.1744511950355698, + "learning_rate": 1.311170977854973e-05, + "loss": 0.4039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40507781505584717, + "step": 6385, + "valid_targets_mean": 28851.4, + "valid_targets_min": 21563 + }, + { + "epoch": 6.819637139807898, + "grad_norm": 0.16895744034595586, + "learning_rate": 1.2996681280675894e-05, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40257859230041504, + "step": 6390, + "valid_targets_mean": 28739.7, + "valid_targets_min": 19245 + }, + { + "epoch": 6.824973319103522, + "grad_norm": 0.177459104380991, + "learning_rate": 1.2882124501078597e-05, + "loss": 0.4001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39968159794807434, + "step": 6395, + "valid_targets_mean": 28733.4, + "valid_targets_min": 20661 + }, + { + "epoch": 6.830309498399147, + "grad_norm": 0.18238957188441435, + "learning_rate": 1.2768040060866593e-05, + "loss": 0.4002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4020434021949768, + "step": 6400, + "valid_targets_mean": 28898.1, + "valid_targets_min": 22863 + }, + { + "epoch": 6.835645677694771, + "grad_norm": 0.15973146615703437, + "learning_rate": 1.2654428578587763e-05, + "loss": 0.4012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40105491876602173, + "step": 6405, + "valid_targets_mean": 28873.5, + "valid_targets_min": 22667 + }, + { + "epoch": 6.840981856990394, + "grad_norm": 0.1798151031695525, + "learning_rate": 1.2541290670225647e-05, + "loss": 0.403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4018270969390869, + "step": 6410, + "valid_targets_mean": 28675.7, + "valid_targets_min": 22289 + }, + { + "epoch": 6.84631803628602, + "grad_norm": 0.16448721303729166, + "learning_rate": 1.2428626949196076e-05, + "loss": 0.4013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4013287425041199, + "step": 6415, + "valid_targets_mean": 28779.9, + "valid_targets_min": 18744 + }, + { + "epoch": 6.851654215581643, + "grad_norm": 0.17415414751858463, + "learning_rate": 1.231643802634409e-05, + "loss": 0.4001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40349093079566956, + "step": 6420, + "valid_targets_mean": 28927.2, + "valid_targets_min": 20736 + }, + { + "epoch": 6.856990394877268, + "grad_norm": 0.16319867986700945, + "learning_rate": 1.220472450994027e-05, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3997553586959839, + "step": 6425, + "valid_targets_mean": 28938.6, + "valid_targets_min": 22659 + }, + { + "epoch": 6.862326574172892, + "grad_norm": 0.16461669022500233, + "learning_rate": 1.2093487005677728e-05, + "loss": 0.3989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3953731656074524, + "step": 6430, + "valid_targets_mean": 28847.4, + "valid_targets_min": 22060 + }, + { + "epoch": 6.867662753468516, + "grad_norm": 0.1724566284023691, + "learning_rate": 1.1982726116668642e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4027242660522461, + "step": 6435, + "valid_targets_mean": 28751.4, + "valid_targets_min": 21767 + }, + { + "epoch": 6.8729989327641405, + "grad_norm": 0.16308302612711748, + "learning_rate": 1.1872442443441113e-05, + "loss": 0.4006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39664605259895325, + "step": 6440, + "valid_targets_mean": 28841.8, + "valid_targets_min": 22914 + }, + { + "epoch": 6.878335112059765, + "grad_norm": 0.1482871997880307, + "learning_rate": 1.1762636583935793e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40228116512298584, + "step": 6445, + "valid_targets_mean": 28875.2, + "valid_targets_min": 20679 + }, + { + "epoch": 6.883671291355389, + "grad_norm": 0.1685029515968394, + "learning_rate": 1.1653309133502777e-05, + "loss": 0.4011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40284842252731323, + "step": 6450, + "valid_targets_mean": 28918.3, + "valid_targets_min": 19134 + }, + { + "epoch": 6.889007470651014, + "grad_norm": 0.15905665248843048, + "learning_rate": 1.1544460684898261e-05, + "loss": 0.4028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39986366033554077, + "step": 6455, + "valid_targets_mean": 28739.9, + "valid_targets_min": 20314 + }, + { + "epoch": 6.894343649946638, + "grad_norm": 0.17853792674427735, + "learning_rate": 1.1436091828281348e-05, + "loss": 0.4006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40273600816726685, + "step": 6460, + "valid_targets_mean": 28903.6, + "valid_targets_min": 20874 + }, + { + "epoch": 6.899679829242262, + "grad_norm": 0.1557539114073379, + "learning_rate": 1.1328203151210915e-05, + "loss": 0.4026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40460121631622314, + "step": 6465, + "valid_targets_mean": 28848.5, + "valid_targets_min": 21189 + }, + { + "epoch": 6.905016008537887, + "grad_norm": 0.16440118079599428, + "learning_rate": 1.1220795238642357e-05, + "loss": 0.3998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3983585834503174, + "step": 6470, + "valid_targets_mean": 28828.2, + "valid_targets_min": 23214 + }, + { + "epoch": 6.910352187833511, + "grad_norm": 0.15390341169540844, + "learning_rate": 1.1113868672924422e-05, + "loss": 0.4003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3979727327823639, + "step": 6475, + "valid_targets_mean": 28908.6, + "valid_targets_min": 22500 + }, + { + "epoch": 6.915688367129135, + "grad_norm": 0.156756752893968, + "learning_rate": 1.1007424033796087e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40027040243148804, + "step": 6480, + "valid_targets_mean": 28882.9, + "valid_targets_min": 20069 + }, + { + "epoch": 6.92102454642476, + "grad_norm": 0.14466757537942485, + "learning_rate": 1.0901461898383424e-05, + "loss": 0.4007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39898619055747986, + "step": 6485, + "valid_targets_mean": 28843.0, + "valid_targets_min": 18406 + }, + { + "epoch": 6.926360725720384, + "grad_norm": 0.16297822304878332, + "learning_rate": 1.079598284119644e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40116146206855774, + "step": 6490, + "valid_targets_mean": 28951.9, + "valid_targets_min": 18291 + }, + { + "epoch": 6.931696905016008, + "grad_norm": 0.17699960140786858, + "learning_rate": 1.06909874341259e-05, + "loss": 0.4016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4010434150695801, + "step": 6495, + "valid_targets_mean": 28819.8, + "valid_targets_min": 21154 + }, + { + "epoch": 6.937033084311633, + "grad_norm": 0.15436844588832435, + "learning_rate": 1.0586476246440403e-05, + "loss": 0.4009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39718085527420044, + "step": 6500, + "valid_targets_mean": 28895.1, + "valid_targets_min": 22361 + }, + { + "epoch": 6.942369263607257, + "grad_norm": 0.15191274323938955, + "learning_rate": 1.0482449844783093e-05, + "loss": 0.4003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40134063363075256, + "step": 6505, + "valid_targets_mean": 28845.7, + "valid_targets_min": 21166 + }, + { + "epoch": 6.9477054429028815, + "grad_norm": 0.16467666322186136, + "learning_rate": 1.0378908793168718e-05, + "loss": 0.4003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40012848377227783, + "step": 6510, + "valid_targets_mean": 28919.9, + "valid_targets_min": 20421 + }, + { + "epoch": 6.953041622198506, + "grad_norm": 0.16302817190639357, + "learning_rate": 1.0275853652980571e-05, + "loss": 0.4017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3995431959629059, + "step": 6515, + "valid_targets_mean": 28810.1, + "valid_targets_min": 20639 + }, + { + "epoch": 6.95837780149413, + "grad_norm": 0.15702759706035485, + "learning_rate": 1.0173284982967335e-05, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3988278806209564, + "step": 6520, + "valid_targets_mean": 28847.2, + "valid_targets_min": 20450 + }, + { + "epoch": 6.963713980789755, + "grad_norm": 0.14701600676799004, + "learning_rate": 1.0071203339240198e-05, + "loss": 0.4019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4036068916320801, + "step": 6525, + "valid_targets_mean": 28974.2, + "valid_targets_min": 22440 + }, + { + "epoch": 6.969050160085379, + "grad_norm": 0.1603744815668825, + "learning_rate": 9.96960927526972e-06, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40067869424819946, + "step": 6530, + "valid_targets_mean": 28709.8, + "valid_targets_min": 16672 + }, + { + "epoch": 6.974386339381003, + "grad_norm": 0.15407947241724276, + "learning_rate": 9.868503341882872e-06, + "loss": 0.4005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3975338935852051, + "step": 6535, + "valid_targets_mean": 28773.6, + "valid_targets_min": 19931 + }, + { + "epoch": 6.979722518676628, + "grad_norm": 0.15866317471533062, + "learning_rate": 9.767886087260125e-06, + "loss": 0.3991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40186864137649536, + "step": 6540, + "valid_targets_mean": 28824.8, + "valid_targets_min": 19711 + }, + { + "epoch": 6.985058697972252, + "grad_norm": 0.1603353701551661, + "learning_rate": 9.667758056932341e-06, + "loss": 0.4001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.398986279964447, + "step": 6545, + "valid_targets_mean": 28972.2, + "valid_targets_min": 22728 + }, + { + "epoch": 6.990394877267876, + "grad_norm": 0.15628933829566188, + "learning_rate": 9.568119793777952e-06, + "loss": 0.4004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.402820885181427, + "step": 6550, + "valid_targets_mean": 28688.5, + "valid_targets_min": 19256 + }, + { + "epoch": 6.995731056563501, + "grad_norm": 0.15171731237911634, + "learning_rate": 9.468971838019869e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3992854356765747, + "step": 6555, + "valid_targets_mean": 28961.2, + "valid_targets_min": 21536 + }, + { + "epoch": 7.001067235859125, + "grad_norm": 0.15330599997099978, + "learning_rate": 9.370314727222718e-06, + "loss": 0.4021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40003207325935364, + "step": 6560, + "valid_targets_mean": 28604.1, + "valid_targets_min": 20828 + }, + { + "epoch": 7.006403415154749, + "grad_norm": 0.16415267947660667, + "learning_rate": 9.272148996289765e-06, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4015958905220032, + "step": 6565, + "valid_targets_mean": 28911.6, + "valid_targets_min": 22307 + }, + { + "epoch": 7.011739594450374, + "grad_norm": 0.15677689118830038, + "learning_rate": 9.174475177460096e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3973429501056671, + "step": 6570, + "valid_targets_mean": 28885.8, + "valid_targets_min": 21403 + }, + { + "epoch": 7.017075773745998, + "grad_norm": 0.1555057933551184, + "learning_rate": 9.077293800305842e-06, + "loss": 0.4007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39989423751831055, + "step": 6575, + "valid_targets_mean": 28826.1, + "valid_targets_min": 21106 + }, + { + "epoch": 7.0224119530416225, + "grad_norm": 0.15198025809307297, + "learning_rate": 8.980605391729035e-06, + "loss": 0.3983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39894890785217285, + "step": 6580, + "valid_targets_mean": 28534.4, + "valid_targets_min": 19891 + }, + { + "epoch": 7.027748132337247, + "grad_norm": 0.14452981870722426, + "learning_rate": 8.884410475959026e-06, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3982957601547241, + "step": 6585, + "valid_targets_mean": 28677.5, + "valid_targets_min": 14588 + }, + { + "epoch": 7.033084311632871, + "grad_norm": 0.15139167888669983, + "learning_rate": 8.788709574549469e-06, + "loss": 0.3964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39983707666397095, + "step": 6590, + "valid_targets_mean": 28721.9, + "valid_targets_min": 20030 + }, + { + "epoch": 7.0384204909284955, + "grad_norm": 0.17697283738640873, + "learning_rate": 8.693503206375531e-06, + "loss": 0.3989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39999592304229736, + "step": 6595, + "valid_targets_mean": 28832.6, + "valid_targets_min": 20481 + }, + { + "epoch": 7.04375667022412, + "grad_norm": 0.1463523461872221, + "learning_rate": 8.598791887631164e-06, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39724496006965637, + "step": 6600, + "valid_targets_mean": 28774.0, + "valid_targets_min": 22752 + }, + { + "epoch": 7.049092849519744, + "grad_norm": 0.14902110723811787, + "learning_rate": 8.504576131826181e-06, + "loss": 0.3992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3978765904903412, + "step": 6605, + "valid_targets_mean": 29010.4, + "valid_targets_min": 22184 + }, + { + "epoch": 7.054429028815369, + "grad_norm": 0.1473453761555798, + "learning_rate": 8.4108564497836e-06, + "loss": 0.3984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3956196904182434, + "step": 6610, + "valid_targets_mean": 28715.8, + "valid_targets_min": 18406 + }, + { + "epoch": 7.059765208110993, + "grad_norm": 0.1550612916087296, + "learning_rate": 8.317633349636712e-06, + "loss": 0.3967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39694520831108093, + "step": 6615, + "valid_targets_mean": 28844.7, + "valid_targets_min": 21374 + }, + { + "epoch": 7.065101387406617, + "grad_norm": 0.1568547835311203, + "learning_rate": 8.224907336826504e-06, + "loss": 0.3987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3986090421676636, + "step": 6620, + "valid_targets_mean": 28702.4, + "valid_targets_min": 19817 + }, + { + "epoch": 7.070437566702241, + "grad_norm": 0.14798714715897823, + "learning_rate": 8.132678914098779e-06, + "loss": 0.3988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39725714921951294, + "step": 6625, + "valid_targets_mean": 28755.3, + "valid_targets_min": 19321 + }, + { + "epoch": 7.075773745997865, + "grad_norm": 0.15140895432312573, + "learning_rate": 8.040948581501495e-06, + "loss": 0.3983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39604485034942627, + "step": 6630, + "valid_targets_mean": 28734.3, + "valid_targets_min": 20528 + }, + { + "epoch": 7.0811099252934895, + "grad_norm": 0.14021405335849116, + "learning_rate": 7.949716836382048e-06, + "loss": 0.3961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3950844407081604, + "step": 6635, + "valid_targets_mean": 28809.1, + "valid_targets_min": 18632 + }, + { + "epoch": 7.086446104589114, + "grad_norm": 0.14555571536553902, + "learning_rate": 7.858984173384598e-06, + "loss": 0.3996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3982555866241455, + "step": 6640, + "valid_targets_mean": 28952.2, + "valid_targets_min": 18491 + }, + { + "epoch": 7.091782283884738, + "grad_norm": 0.14723295242994291, + "learning_rate": 7.768751084447357e-06, + "loss": 0.3967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3947444558143616, + "step": 6645, + "valid_targets_mean": 28932.8, + "valid_targets_min": 22607 + }, + { + "epoch": 7.0971184631803625, + "grad_norm": 0.15320077756778971, + "learning_rate": 7.67901805879987e-06, + "loss": 0.3968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39684662222862244, + "step": 6650, + "valid_targets_mean": 28865.5, + "valid_targets_min": 23157 + }, + { + "epoch": 7.102454642475987, + "grad_norm": 0.1544152537510823, + "learning_rate": 7.589785582960441e-06, + "loss": 0.3981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39948925375938416, + "step": 6655, + "valid_targets_mean": 28771.1, + "valid_targets_min": 18799 + }, + { + "epoch": 7.107790821771611, + "grad_norm": 0.14117398692965866, + "learning_rate": 7.501054140733521e-06, + "loss": 0.398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39882612228393555, + "step": 6660, + "valid_targets_mean": 28572.9, + "valid_targets_min": 20299 + }, + { + "epoch": 7.113127001067236, + "grad_norm": 0.16295455416882793, + "learning_rate": 7.412824213206981e-06, + "loss": 0.3973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39467161893844604, + "step": 6665, + "valid_targets_mean": 28756.4, + "valid_targets_min": 19687 + }, + { + "epoch": 7.11846318036286, + "grad_norm": 0.15492019017386385, + "learning_rate": 7.325096278749599e-06, + "loss": 0.3957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39982277154922485, + "step": 6670, + "valid_targets_mean": 28843.5, + "valid_targets_min": 21406 + }, + { + "epoch": 7.123799359658484, + "grad_norm": 0.14707840331415228, + "learning_rate": 7.237870813008362e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39708107709884644, + "step": 6675, + "valid_targets_mean": 28789.7, + "valid_targets_min": 19764 + }, + { + "epoch": 7.129135538954109, + "grad_norm": 0.14771471386609525, + "learning_rate": 7.151148288906051e-06, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39893800020217896, + "step": 6680, + "valid_targets_mean": 28768.1, + "valid_targets_min": 20127 + }, + { + "epoch": 7.134471718249733, + "grad_norm": 0.14680754081055059, + "learning_rate": 7.064929176638469e-06, + "loss": 0.3967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3978503346443176, + "step": 6685, + "valid_targets_mean": 28842.6, + "valid_targets_min": 22315 + }, + { + "epoch": 7.139807897545357, + "grad_norm": 0.14806307005666994, + "learning_rate": 6.979213943672058e-06, + "loss": 0.3987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40188246965408325, + "step": 6690, + "valid_targets_mean": 28730.9, + "valid_targets_min": 18655 + }, + { + "epoch": 7.145144076840982, + "grad_norm": 0.14834292296077992, + "learning_rate": 6.8940030547413044e-06, + "loss": 0.3974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39668893814086914, + "step": 6695, + "valid_targets_mean": 28693.6, + "valid_targets_min": 21590 + }, + { + "epoch": 7.150480256136606, + "grad_norm": 0.14362800777994875, + "learning_rate": 6.809296971846213e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39519941806793213, + "step": 6700, + "valid_targets_mean": 28812.1, + "valid_targets_min": 22316 + }, + { + "epoch": 7.1558164354322304, + "grad_norm": 0.1769964892974021, + "learning_rate": 6.725096154249822e-06, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39827579259872437, + "step": 6705, + "valid_targets_mean": 28764.0, + "valid_targets_min": 21474 + }, + { + "epoch": 7.161152614727855, + "grad_norm": 0.14440606541634426, + "learning_rate": 6.6414010584756826e-06, + "loss": 0.3968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3953492343425751, + "step": 6710, + "valid_targets_mean": 28923.6, + "valid_targets_min": 23038 + }, + { + "epoch": 7.166488794023479, + "grad_norm": 0.14340621091725506, + "learning_rate": 6.558212138305375e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3969920873641968, + "step": 6715, + "valid_targets_mean": 28838.3, + "valid_targets_min": 22743 + }, + { + "epoch": 7.1718249733191035, + "grad_norm": 0.1390070556862948, + "learning_rate": 6.475529844776163e-06, + "loss": 0.3954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3945442736148834, + "step": 6720, + "valid_targets_mean": 28672.8, + "valid_targets_min": 17816 + }, + { + "epoch": 7.177161152614728, + "grad_norm": 0.13867783752463655, + "learning_rate": 6.3933546261783496e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3949880301952362, + "step": 6725, + "valid_targets_mean": 28804.1, + "valid_targets_min": 15165 + }, + { + "epoch": 7.182497331910352, + "grad_norm": 0.14234016351718132, + "learning_rate": 6.3116869280530514e-06, + "loss": 0.3986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3978882133960724, + "step": 6730, + "valid_targets_mean": 28754.6, + "valid_targets_min": 19396 + }, + { + "epoch": 7.187833511205977, + "grad_norm": 0.13609739413508615, + "learning_rate": 6.230527193189639e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39371705055236816, + "step": 6735, + "valid_targets_mean": 29004.2, + "valid_targets_min": 21072 + }, + { + "epoch": 7.193169690501601, + "grad_norm": 0.14313698108119827, + "learning_rate": 6.149875861623411e-06, + "loss": 0.3968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3963935077190399, + "step": 6740, + "valid_targets_mean": 28970.4, + "valid_targets_min": 22776 + }, + { + "epoch": 7.198505869797225, + "grad_norm": 0.17382222681898493, + "learning_rate": 6.069733370633168e-06, + "loss": 0.3986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4010206460952759, + "step": 6745, + "valid_targets_mean": 28954.6, + "valid_targets_min": 22455 + }, + { + "epoch": 7.20384204909285, + "grad_norm": 0.13672333265107187, + "learning_rate": 5.990100154738842e-06, + "loss": 0.3974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3966425955295563, + "step": 6750, + "valid_targets_mean": 28855.2, + "valid_targets_min": 19999 + }, + { + "epoch": 7.209178228388474, + "grad_norm": 0.14133605489320078, + "learning_rate": 5.910976645699174e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3986111879348755, + "step": 6755, + "valid_targets_mean": 28904.9, + "valid_targets_min": 20203 + }, + { + "epoch": 7.214514407684098, + "grad_norm": 0.13676766532037, + "learning_rate": 5.832363272509378e-06, + "loss": 0.3982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3987554907798767, + "step": 6760, + "valid_targets_mean": 28837.3, + "valid_targets_min": 22529 + }, + { + "epoch": 7.219850586979723, + "grad_norm": 0.1407521511478025, + "learning_rate": 5.754260461398764e-06, + "loss": 0.3975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40251851081848145, + "step": 6765, + "valid_targets_mean": 28777.5, + "valid_targets_min": 17618 + }, + { + "epoch": 7.225186766275347, + "grad_norm": 0.14439183596932256, + "learning_rate": 5.6766686358284525e-06, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40174761414527893, + "step": 6770, + "valid_targets_mean": 28680.2, + "valid_targets_min": 22633 + }, + { + "epoch": 7.230522945570971, + "grad_norm": 0.14054950386320145, + "learning_rate": 5.599588216489071e-06, + "loss": 0.4004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39928752183914185, + "step": 6775, + "valid_targets_mean": 28880.4, + "valid_targets_min": 22468 + }, + { + "epoch": 7.235859124866596, + "grad_norm": 0.14264311473697092, + "learning_rate": 5.523019621298542e-06, + "loss": 0.3969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39656123518943787, + "step": 6780, + "valid_targets_mean": 28882.2, + "valid_targets_min": 22228 + }, + { + "epoch": 7.24119530416222, + "grad_norm": 0.13433578292405557, + "learning_rate": 5.446963265399674e-06, + "loss": 0.3982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39557063579559326, + "step": 6785, + "valid_targets_mean": 28956.7, + "valid_targets_min": 22959 + }, + { + "epoch": 7.2465314834578445, + "grad_norm": 0.1407962413902608, + "learning_rate": 5.371419561158042e-06, + "loss": 0.3986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4022854268550873, + "step": 6790, + "valid_targets_mean": 28571.2, + "valid_targets_min": 14549 + }, + { + "epoch": 7.251867662753469, + "grad_norm": 0.14403343716608968, + "learning_rate": 5.296388918159723e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39685022830963135, + "step": 6795, + "valid_targets_mean": 28883.9, + "valid_targets_min": 20478 + }, + { + "epoch": 7.257203842049093, + "grad_norm": 0.14013211142564314, + "learning_rate": 5.221871743209039e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3962429165840149, + "step": 6800, + "valid_targets_mean": 28686.2, + "valid_targets_min": 21420 + }, + { + "epoch": 7.2625400213447175, + "grad_norm": 0.13661800280592126, + "learning_rate": 5.1478684403263425e-06, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3977741003036499, + "step": 6805, + "valid_targets_mean": 28841.9, + "valid_targets_min": 21571 + }, + { + "epoch": 7.267876200640342, + "grad_norm": 0.13806049467788378, + "learning_rate": 5.0743794107458664e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3959900140762329, + "step": 6810, + "valid_targets_mean": 28860.0, + "valid_targets_min": 18804 + }, + { + "epoch": 7.273212379935966, + "grad_norm": 0.13400193213805908, + "learning_rate": 5.001405052913577e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3986443877220154, + "step": 6815, + "valid_targets_mean": 28876.5, + "valid_targets_min": 22960 + }, + { + "epoch": 7.27854855923159, + "grad_norm": 0.1318872747199963, + "learning_rate": 4.9289457624848714e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40434730052948, + "step": 6820, + "valid_targets_mean": 28844.1, + "valid_targets_min": 17444 + }, + { + "epoch": 7.283884738527215, + "grad_norm": 0.13980805664155702, + "learning_rate": 4.85700193232268e-06, + "loss": 0.3953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39328184723854065, + "step": 6825, + "valid_targets_mean": 28844.9, + "valid_targets_min": 21078 + }, + { + "epoch": 7.289220917822838, + "grad_norm": 0.14814865784308243, + "learning_rate": 4.785573952495059e-06, + "loss": 0.3986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39459824562072754, + "step": 6830, + "valid_targets_mean": 28762.5, + "valid_targets_min": 21269 + }, + { + "epoch": 7.294557097118463, + "grad_norm": 0.1477617513226932, + "learning_rate": 4.714662210273246e-06, + "loss": 0.3998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.398124635219574, + "step": 6835, + "valid_targets_mean": 28962.0, + "valid_targets_min": 20359 + }, + { + "epoch": 7.299893276414087, + "grad_norm": 0.13491286518951615, + "learning_rate": 4.644267090129562e-06, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39378926157951355, + "step": 6840, + "valid_targets_mean": 28842.9, + "valid_targets_min": 22195 + }, + { + "epoch": 7.3052294557097115, + "grad_norm": 0.13590823625556828, + "learning_rate": 4.5743889737352215e-06, + "loss": 0.3987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39613455533981323, + "step": 6845, + "valid_targets_mean": 28674.6, + "valid_targets_min": 20348 + }, + { + "epoch": 7.310565635005336, + "grad_norm": 0.14015434698468512, + "learning_rate": 4.505028239958353e-06, + "loss": 0.3948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39379096031188965, + "step": 6850, + "valid_targets_mean": 28751.0, + "valid_targets_min": 16576 + }, + { + "epoch": 7.31590181430096, + "grad_norm": 0.1375274836111597, + "learning_rate": 4.436185264861936e-06, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39162224531173706, + "step": 6855, + "valid_targets_mean": 28707.3, + "valid_targets_min": 19748 + }, + { + "epoch": 7.321237993596585, + "grad_norm": 0.15924018291017472, + "learning_rate": 4.367860421701731e-06, + "loss": 0.3959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3961021602153778, + "step": 6860, + "valid_targets_mean": 28817.5, + "valid_targets_min": 21668 + }, + { + "epoch": 7.326574172892209, + "grad_norm": 0.13253022328531194, + "learning_rate": 4.300054080924254e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3973386287689209, + "step": 6865, + "valid_targets_mean": 28927.5, + "valid_targets_min": 18508 + }, + { + "epoch": 7.331910352187833, + "grad_norm": 0.13692155487606714, + "learning_rate": 4.2327666101647735e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39944925904273987, + "step": 6870, + "valid_targets_mean": 28797.9, + "valid_targets_min": 20510 + }, + { + "epoch": 7.337246531483458, + "grad_norm": 0.13264873336052005, + "learning_rate": 4.165998374245395e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3951477110385895, + "step": 6875, + "valid_targets_mean": 28752.1, + "valid_targets_min": 17730 + }, + { + "epoch": 7.342582710779082, + "grad_norm": 0.14478248166830496, + "learning_rate": 4.099749735172909e-06, + "loss": 0.3989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3972729444503784, + "step": 6880, + "valid_targets_mean": 28750.5, + "valid_targets_min": 22412 + }, + { + "epoch": 7.347918890074706, + "grad_norm": 0.1385750912506846, + "learning_rate": 4.034021052137027e-06, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3962293565273285, + "step": 6885, + "valid_targets_mean": 28836.2, + "valid_targets_min": 20189 + }, + { + "epoch": 7.353255069370331, + "grad_norm": 0.14214062049125484, + "learning_rate": 3.968812681508316e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39318692684173584, + "step": 6890, + "valid_targets_mean": 28721.9, + "valid_targets_min": 21114 + }, + { + "epoch": 7.358591248665955, + "grad_norm": 0.1338862866468272, + "learning_rate": 3.904124976836254e-06, + "loss": 0.3984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3970756232738495, + "step": 6895, + "valid_targets_mean": 28713.5, + "valid_targets_min": 19112 + }, + { + "epoch": 7.363927427961579, + "grad_norm": 0.14940392657549093, + "learning_rate": 3.8399582888473895e-06, + "loss": 0.3958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3941649794578552, + "step": 6900, + "valid_targets_mean": 28974.6, + "valid_targets_min": 20549 + }, + { + "epoch": 7.369263607257204, + "grad_norm": 0.13475410359357387, + "learning_rate": 3.776312965443374e-06, + "loss": 0.3972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3961244821548462, + "step": 6905, + "valid_targets_mean": 28744.7, + "valid_targets_min": 20562 + }, + { + "epoch": 7.374599786552828, + "grad_norm": 0.13723555168541482, + "learning_rate": 3.713189351699131e-06, + "loss": 0.3981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40024417638778687, + "step": 6910, + "valid_targets_mean": 28850.6, + "valid_targets_min": 21788 + }, + { + "epoch": 7.3799359658484525, + "grad_norm": 0.1332438575036423, + "learning_rate": 3.6505877898609263e-06, + "loss": 0.3975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.399570494890213, + "step": 6915, + "valid_targets_mean": 28985.8, + "valid_targets_min": 22477 + }, + { + "epoch": 7.385272145144077, + "grad_norm": 0.13848205951792947, + "learning_rate": 3.5885086193445883e-06, + "loss": 0.3964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3938809633255005, + "step": 6920, + "valid_targets_mean": 28860.3, + "valid_targets_min": 22807 + }, + { + "epoch": 7.390608324439701, + "grad_norm": 0.13650005467831897, + "learning_rate": 3.526952176733578e-06, + "loss": 0.3948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39737313985824585, + "step": 6925, + "valid_targets_mean": 28816.6, + "valid_targets_min": 21229 + }, + { + "epoch": 7.3959445037353255, + "grad_norm": 0.12888511781642153, + "learning_rate": 3.46591879577719e-06, + "loss": 0.3977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3936184048652649, + "step": 6930, + "valid_targets_mean": 28959.3, + "valid_targets_min": 17707 + }, + { + "epoch": 7.40128068303095, + "grad_norm": 0.13263492604068003, + "learning_rate": 3.4054088073888435e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3939823508262634, + "step": 6935, + "valid_targets_mean": 28700.5, + "valid_targets_min": 21618 + }, + { + "epoch": 7.406616862326574, + "grad_norm": 0.13102061046697075, + "learning_rate": 3.3454225396441164e-06, + "loss": 0.3973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.397549033164978, + "step": 6940, + "valid_targets_mean": 28769.5, + "valid_targets_min": 20988 + }, + { + "epoch": 7.411953041622199, + "grad_norm": 0.13206284597273021, + "learning_rate": 3.285960317779102e-06, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39504343271255493, + "step": 6945, + "valid_targets_mean": 28798.3, + "valid_targets_min": 19081 + }, + { + "epoch": 7.417289220917823, + "grad_norm": 0.14247388308189723, + "learning_rate": 3.2270224641886117e-06, + "loss": 0.3998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39570754766464233, + "step": 6950, + "valid_targets_mean": 28884.1, + "valid_targets_min": 21183 + }, + { + "epoch": 7.422625400213447, + "grad_norm": 0.12933879296966458, + "learning_rate": 3.168609298424374e-06, + "loss": 0.3941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39268964529037476, + "step": 6955, + "valid_targets_mean": 28773.8, + "valid_targets_min": 21747 + }, + { + "epoch": 7.427961579509072, + "grad_norm": 0.13258932817271468, + "learning_rate": 3.110721137193384e-06, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3964608907699585, + "step": 6960, + "valid_targets_mean": 28910.0, + "valid_targets_min": 21896 + }, + { + "epoch": 7.433297758804696, + "grad_norm": 0.14054028295501642, + "learning_rate": 3.053358294356101e-06, + "loss": 0.3953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39511436223983765, + "step": 6965, + "valid_targets_mean": 28822.8, + "valid_targets_min": 22606 + }, + { + "epoch": 7.43863393810032, + "grad_norm": 0.1447540553523416, + "learning_rate": 2.9965210809248412e-06, + "loss": 0.3988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3964189887046814, + "step": 6970, + "valid_targets_mean": 28971.0, + "valid_targets_min": 20946 + }, + { + "epoch": 7.443970117395945, + "grad_norm": 0.12916593251265385, + "learning_rate": 2.9402098050619775e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3927098512649536, + "step": 6975, + "valid_targets_mean": 29101.2, + "valid_targets_min": 20495 + }, + { + "epoch": 7.449306296691569, + "grad_norm": 0.15834400094498974, + "learning_rate": 2.884424772078398e-06, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39576098322868347, + "step": 6980, + "valid_targets_mean": 28813.8, + "valid_targets_min": 21778 + }, + { + "epoch": 7.454642475987193, + "grad_norm": 0.13571641243628318, + "learning_rate": 2.829166284431761e-06, + "loss": 0.3992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3991106152534485, + "step": 6985, + "valid_targets_mean": 28901.6, + "valid_targets_min": 19732 + }, + { + "epoch": 7.459978655282818, + "grad_norm": 0.1331103948023071, + "learning_rate": 2.7744346417248546e-06, + "loss": 0.3974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3965522050857544, + "step": 6990, + "valid_targets_mean": 28912.5, + "valid_targets_min": 17680 + }, + { + "epoch": 7.465314834578442, + "grad_norm": 0.1319675964651442, + "learning_rate": 2.7202301407040274e-06, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3990844488143921, + "step": 6995, + "valid_targets_mean": 28948.4, + "valid_targets_min": 19629 + }, + { + "epoch": 7.4706510138740665, + "grad_norm": 0.1405373888533213, + "learning_rate": 2.6665530752575274e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3977741599082947, + "step": 7000, + "valid_targets_mean": 28642.5, + "valid_targets_min": 19049 + }, + { + "epoch": 7.475987193169691, + "grad_norm": 0.12816007543103924, + "learning_rate": 2.613403736413944e-06, + "loss": 0.3982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3968815803527832, + "step": 7005, + "valid_targets_mean": 28858.9, + "valid_targets_min": 22519 + }, + { + "epoch": 7.481323372465315, + "grad_norm": 0.13763866771563157, + "learning_rate": 2.560782412340623e-06, + "loss": 0.398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3996983766555786, + "step": 7010, + "valid_targets_mean": 28726.6, + "valid_targets_min": 19324 + }, + { + "epoch": 7.48665955176094, + "grad_norm": 0.13402555031487307, + "learning_rate": 2.508689388342078e-06, + "loss": 0.3973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4007349908351898, + "step": 7015, + "valid_targets_mean": 28784.0, + "valid_targets_min": 20900 + }, + { + "epoch": 7.491995731056564, + "grad_norm": 0.13388666886846984, + "learning_rate": 2.45712494685848e-06, + "loss": 0.3977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.399408757686615, + "step": 7020, + "valid_targets_mean": 28943.9, + "valid_targets_min": 18428 + }, + { + "epoch": 7.497331910352187, + "grad_norm": 0.13471132033960173, + "learning_rate": 2.4060893674640815e-06, + "loss": 0.398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4004710912704468, + "step": 7025, + "valid_targets_mean": 28850.0, + "valid_targets_min": 22619 + }, + { + "epoch": 7.502668089647813, + "grad_norm": 0.12545470797822975, + "learning_rate": 2.3555829268657626e-06, + "loss": 0.398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4004118740558624, + "step": 7030, + "valid_targets_mean": 28785.9, + "valid_targets_min": 21242 + }, + { + "epoch": 7.508004268943436, + "grad_norm": 0.1309840290867499, + "learning_rate": 2.305605898901486e-06, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.399819016456604, + "step": 7035, + "valid_targets_mean": 29010.9, + "valid_targets_min": 16563 + }, + { + "epoch": 7.513340448239061, + "grad_norm": 0.12977304528913586, + "learning_rate": 2.25615855453879e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39655691385269165, + "step": 7040, + "valid_targets_mean": 28872.3, + "valid_targets_min": 22548 + }, + { + "epoch": 7.518676627534685, + "grad_norm": 0.13233513610791645, + "learning_rate": 2.207241161873419e-06, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39584222435951233, + "step": 7045, + "valid_targets_mean": 28785.3, + "valid_targets_min": 22175 + }, + { + "epoch": 7.524012806830309, + "grad_norm": 0.1306727802585294, + "learning_rate": 2.15885398612774e-06, + "loss": 0.3943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39712512493133545, + "step": 7050, + "valid_targets_mean": 28708.5, + "valid_targets_min": 19500 + }, + { + "epoch": 7.5293489861259335, + "grad_norm": 0.14629483828482254, + "learning_rate": 2.110997289649397e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39369213581085205, + "step": 7055, + "valid_targets_mean": 28865.4, + "valid_targets_min": 17438 + }, + { + "epoch": 7.534685165421558, + "grad_norm": 0.13031826983551667, + "learning_rate": 2.0636713319098444e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39837566018104553, + "step": 7060, + "valid_targets_mean": 28751.2, + "valid_targets_min": 20910 + }, + { + "epoch": 7.540021344717182, + "grad_norm": 0.13010925145229557, + "learning_rate": 2.0168763695029735e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39550474286079407, + "step": 7065, + "valid_targets_mean": 28783.4, + "valid_targets_min": 20371 + }, + { + "epoch": 7.545357524012807, + "grad_norm": 0.13859542215077358, + "learning_rate": 1.9706126561436775e-06, + "loss": 0.3993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3975289762020111, + "step": 7070, + "valid_targets_mean": 28944.4, + "valid_targets_min": 17638 + }, + { + "epoch": 7.550693703308431, + "grad_norm": 0.12883914997866086, + "learning_rate": 1.9248804426665323e-06, + "loss": 0.3963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.395477831363678, + "step": 7075, + "valid_targets_mean": 28890.7, + "valid_targets_min": 22324 + }, + { + "epoch": 7.556029882604055, + "grad_norm": 0.12680204603676157, + "learning_rate": 1.8796799770243845e-06, + "loss": 0.3963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39461207389831543, + "step": 7080, + "valid_targets_mean": 28871.8, + "valid_targets_min": 21378 + }, + { + "epoch": 7.56136606189968, + "grad_norm": 0.12948065308048, + "learning_rate": 1.835011504287032e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3954583406448364, + "step": 7085, + "valid_targets_mean": 28856.8, + "valid_targets_min": 21181 + }, + { + "epoch": 7.566702241195304, + "grad_norm": 0.12563199927845528, + "learning_rate": 1.790875266639891e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39542630314826965, + "step": 7090, + "valid_targets_mean": 28895.1, + "valid_targets_min": 18822 + }, + { + "epoch": 7.572038420490928, + "grad_norm": 0.13220859394541173, + "learning_rate": 1.7472715033826747e-06, + "loss": 0.396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39799848198890686, + "step": 7095, + "valid_targets_mean": 28812.4, + "valid_targets_min": 20946 + }, + { + "epoch": 7.577374599786553, + "grad_norm": 0.12830894957199165, + "learning_rate": 1.7042004509281284e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3940737843513489, + "step": 7100, + "valid_targets_mean": 28974.1, + "valid_targets_min": 22172 + }, + { + "epoch": 7.582710779082177, + "grad_norm": 0.12304144890226744, + "learning_rate": 1.6616623428006961e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40141934156417847, + "step": 7105, + "valid_targets_mean": 28829.5, + "valid_targets_min": 21348 + }, + { + "epoch": 7.588046958377801, + "grad_norm": 0.13685029611412847, + "learning_rate": 1.6196574096353e-06, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3940128982067108, + "step": 7110, + "valid_targets_mean": 28937.2, + "valid_targets_min": 22069 + }, + { + "epoch": 7.593383137673426, + "grad_norm": 0.1320911496723597, + "learning_rate": 1.578185879176064e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40384820103645325, + "step": 7115, + "valid_targets_mean": 28760.0, + "valid_targets_min": 21903 + }, + { + "epoch": 7.59871931696905, + "grad_norm": 0.12524210991399579, + "learning_rate": 1.5372479762750691e-06, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39434826374053955, + "step": 7120, + "valid_targets_mean": 28681.2, + "valid_targets_min": 21813 + }, + { + "epoch": 7.6040554962646745, + "grad_norm": 0.12842413602049793, + "learning_rate": 1.4968439228912e-06, + "loss": 0.3983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3997383415699005, + "step": 7125, + "valid_targets_mean": 28874.6, + "valid_targets_min": 22386 + }, + { + "epoch": 7.609391675560299, + "grad_norm": 0.13435131545885384, + "learning_rate": 1.4569739380888458e-06, + "loss": 0.3972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4004368484020233, + "step": 7130, + "valid_targets_mean": 28933.9, + "valid_targets_min": 19047 + }, + { + "epoch": 7.614727854855923, + "grad_norm": 0.1327719002764592, + "learning_rate": 1.4176382380367782e-06, + "loss": 0.3994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39994803071022034, + "step": 7135, + "valid_targets_mean": 28925.2, + "valid_targets_min": 21066 + }, + { + "epoch": 7.6200640341515475, + "grad_norm": 0.1299401408543429, + "learning_rate": 1.3788370360069636e-06, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3979865312576294, + "step": 7140, + "valid_targets_mean": 28861.9, + "valid_targets_min": 23294 + }, + { + "epoch": 7.625400213447172, + "grad_norm": 0.15674138969926396, + "learning_rate": 1.3405705423734094e-06, + "loss": 0.396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3987683653831482, + "step": 7145, + "valid_targets_mean": 28867.2, + "valid_targets_min": 21800 + }, + { + "epoch": 7.630736392742796, + "grad_norm": 0.12468537664092988, + "learning_rate": 1.3028389646110084e-06, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39801982045173645, + "step": 7150, + "valid_targets_mean": 28899.1, + "valid_targets_min": 19178 + }, + { + "epoch": 7.636072572038421, + "grad_norm": 0.12518680881130087, + "learning_rate": 1.2656425072944068e-06, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4001878499984741, + "step": 7155, + "valid_targets_mean": 28741.1, + "valid_targets_min": 20442 + }, + { + "epoch": 7.641408751334045, + "grad_norm": 0.13378082935458932, + "learning_rate": 1.228981372096949e-06, + "loss": 0.3954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3943834602832794, + "step": 7160, + "valid_targets_mean": 28739.6, + "valid_targets_min": 21336 + }, + { + "epoch": 7.646744930629669, + "grad_norm": 0.12480255974177112, + "learning_rate": 1.192855757789546e-06, + "loss": 0.396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39438241720199585, + "step": 7165, + "valid_targets_mean": 28988.1, + "valid_targets_min": 22430 + }, + { + "epoch": 7.652081109925294, + "grad_norm": 0.12332588429332686, + "learning_rate": 1.1572658602395426e-06, + "loss": 0.3987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40162521600723267, + "step": 7170, + "valid_targets_mean": 28919.3, + "valid_targets_min": 23571 + }, + { + "epoch": 7.657417289220918, + "grad_norm": 0.1271716835328731, + "learning_rate": 1.122211872409784e-06, + "loss": 0.3987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3963414430618286, + "step": 7175, + "valid_targets_mean": 28940.8, + "valid_targets_min": 22922 + }, + { + "epoch": 7.662753468516542, + "grad_norm": 0.13031537304464122, + "learning_rate": 1.087693984357452e-06, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39790183305740356, + "step": 7180, + "valid_targets_mean": 28825.0, + "valid_targets_min": 19505 + }, + { + "epoch": 7.668089647812167, + "grad_norm": 0.15005193867568728, + "learning_rate": 1.0537123832330964e-06, + "loss": 0.3976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39635568857192993, + "step": 7185, + "valid_targets_mean": 28871.7, + "valid_targets_min": 20854 + }, + { + "epoch": 7.673425827107791, + "grad_norm": 0.12521205270099878, + "learning_rate": 1.0202672532796055e-06, + "loss": 0.3977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.397771418094635, + "step": 7190, + "valid_targets_mean": 28996.4, + "valid_targets_min": 18700 + }, + { + "epoch": 7.678762006403415, + "grad_norm": 0.1263847601133633, + "learning_rate": 9.87358775831182e-07, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39790722727775574, + "step": 7195, + "valid_targets_mean": 28948.4, + "valid_targets_min": 22960 + }, + { + "epoch": 7.68409818569904, + "grad_norm": 0.14131809458762648, + "learning_rate": 9.549871293124234e-07, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3980499804019928, + "step": 7200, + "valid_targets_mean": 28758.7, + "valid_targets_min": 20945 + }, + { + "epoch": 7.689434364994664, + "grad_norm": 0.1253039153817784, + "learning_rate": 9.231524892372889e-07, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3978698253631592, + "step": 7205, + "valid_targets_mean": 28422.1, + "valid_targets_min": 17803 + }, + { + "epoch": 7.6947705442902885, + "grad_norm": 0.12140794592880637, + "learning_rate": 8.918550282081772e-07, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39450564980506897, + "step": 7210, + "valid_targets_mean": 28683.9, + "valid_targets_min": 20870 + }, + { + "epoch": 7.700106723585913, + "grad_norm": 0.12116805708832058, + "learning_rate": 8.610949159149728e-07, + "loss": 0.3961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39463678002357483, + "step": 7215, + "valid_targets_mean": 28825.8, + "valid_targets_min": 19874 + }, + { + "epoch": 7.705442902881536, + "grad_norm": 0.12281107442757595, + "learning_rate": 8.308723191341683e-07, + "loss": 0.3967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39416587352752686, + "step": 7220, + "valid_targets_mean": 28682.7, + "valid_targets_min": 18031 + }, + { + "epoch": 7.710779082177162, + "grad_norm": 0.14860605326149026, + "learning_rate": 8.011874017279208e-07, + "loss": 0.3972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3981039822101593, + "step": 7225, + "valid_targets_mean": 28751.4, + "valid_targets_min": 21110 + }, + { + "epoch": 7.716115261472785, + "grad_norm": 0.12723855024397449, + "learning_rate": 7.72040324643164e-07, + "loss": 0.3977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39874446392059326, + "step": 7230, + "valid_targets_mean": 28755.7, + "valid_targets_min": 22233 + }, + { + "epoch": 7.72145144076841, + "grad_norm": 0.11917391741979977, + "learning_rate": 7.434312459107639e-07, + "loss": 0.3994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40069904923439026, + "step": 7235, + "valid_targets_mean": 28949.6, + "valid_targets_min": 20294 + }, + { + "epoch": 7.726787620064034, + "grad_norm": 0.12341547556245404, + "learning_rate": 7.153603206446203e-07, + "loss": 0.3975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39813292026519775, + "step": 7240, + "valid_targets_mean": 28706.5, + "valid_targets_min": 16608 + }, + { + "epoch": 7.732123799359658, + "grad_norm": 0.14878463296184147, + "learning_rate": 6.878277010408774e-07, + "loss": 0.3963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39200353622436523, + "step": 7245, + "valid_targets_mean": 28828.8, + "valid_targets_min": 21907 + }, + { + "epoch": 7.7374599786552825, + "grad_norm": 0.1221577876639434, + "learning_rate": 6.60833536377059e-07, + "loss": 0.3975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4019637107849121, + "step": 7250, + "valid_targets_mean": 28878.7, + "valid_targets_min": 18154 + }, + { + "epoch": 7.742796157950907, + "grad_norm": 0.12393055640085526, + "learning_rate": 6.343779730112464e-07, + "loss": 0.3973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3989558815956116, + "step": 7255, + "valid_targets_mean": 28855.9, + "valid_targets_min": 23389 + }, + { + "epoch": 7.748132337246531, + "grad_norm": 0.12330587831168091, + "learning_rate": 6.084611543813679e-07, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40006476640701294, + "step": 7260, + "valid_targets_mean": 28777.5, + "valid_targets_min": 19307 + }, + { + "epoch": 7.7534685165421555, + "grad_norm": 0.12246315853862788, + "learning_rate": 5.830832210042991e-07, + "loss": 0.396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3977513909339905, + "step": 7265, + "valid_targets_mean": 28741.4, + "valid_targets_min": 22047 + }, + { + "epoch": 7.75880469583778, + "grad_norm": 0.1247950006517304, + "learning_rate": 5.582443104752089e-07, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3957824409008026, + "step": 7270, + "valid_targets_mean": 28763.2, + "valid_targets_min": 18758 + }, + { + "epoch": 7.764140875133404, + "grad_norm": 0.12264806306956744, + "learning_rate": 5.339445574667479e-07, + "loss": 0.3981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39586859941482544, + "step": 7275, + "valid_targets_mean": 28837.8, + "valid_targets_min": 15054 + }, + { + "epoch": 7.769477054429029, + "grad_norm": 0.14603501645455727, + "learning_rate": 5.101840937283498e-07, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3978617787361145, + "step": 7280, + "valid_targets_mean": 28967.5, + "valid_targets_min": 18255 + }, + { + "epoch": 7.774813233724653, + "grad_norm": 0.12172511207399195, + "learning_rate": 4.869630480854981e-07, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3975546658039093, + "step": 7285, + "valid_targets_mean": 29052.4, + "valid_targets_min": 21218 + }, + { + "epoch": 7.780149413020277, + "grad_norm": 0.12261141364651407, + "learning_rate": 4.6428154643904933e-07, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39828208088874817, + "step": 7290, + "valid_targets_mean": 28853.3, + "valid_targets_min": 20879 + }, + { + "epoch": 7.785485592315902, + "grad_norm": 0.11811955548665705, + "learning_rate": 4.4213971176451095e-07, + "loss": 0.3938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3930383324623108, + "step": 7295, + "valid_targets_mean": 28826.9, + "valid_targets_min": 21494 + }, + { + "epoch": 7.790821771611526, + "grad_norm": 0.12841526707703033, + "learning_rate": 4.2053766411144225e-07, + "loss": 0.3985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39577236771583557, + "step": 7300, + "valid_targets_mean": 28826.0, + "valid_targets_min": 22310 + }, + { + "epoch": 7.79615795090715, + "grad_norm": 0.12011802895684066, + "learning_rate": 3.994755206027212e-07, + "loss": 0.3956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39808082580566406, + "step": 7305, + "valid_targets_mean": 28803.4, + "valid_targets_min": 21407 + }, + { + "epoch": 7.801494130202775, + "grad_norm": 0.12010863458140218, + "learning_rate": 3.789533954339897e-07, + "loss": 0.3983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3978669047355652, + "step": 7310, + "valid_targets_mean": 28767.3, + "valid_targets_min": 19461 + }, + { + "epoch": 7.806830309498399, + "grad_norm": 0.12106939890963439, + "learning_rate": 3.589713998729427e-07, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3986453115940094, + "step": 7315, + "valid_targets_mean": 28953.8, + "valid_targets_min": 23091 + }, + { + "epoch": 7.812166488794023, + "grad_norm": 0.14095560238184315, + "learning_rate": 3.395296422588512e-07, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39821505546569824, + "step": 7320, + "valid_targets_mean": 28776.2, + "valid_targets_min": 20422 + }, + { + "epoch": 7.817502668089648, + "grad_norm": 0.1215992441103773, + "learning_rate": 3.20628228001818e-07, + "loss": 0.3974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4034339189529419, + "step": 7325, + "valid_targets_mean": 28778.5, + "valid_targets_min": 18860 + }, + { + "epoch": 7.822838847385272, + "grad_norm": 0.12213739439146389, + "learning_rate": 3.022672595823672e-07, + "loss": 0.3966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3980867266654968, + "step": 7330, + "valid_targets_mean": 28801.7, + "valid_targets_min": 20663 + }, + { + "epoch": 7.8281750266808965, + "grad_norm": 0.11906318514646144, + "learning_rate": 2.844468365507447e-07, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39695197343826294, + "step": 7335, + "valid_targets_mean": 28957.1, + "valid_targets_min": 21597 + }, + { + "epoch": 7.833511205976521, + "grad_norm": 0.11748452176757196, + "learning_rate": 2.6716705552649603e-07, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3960111141204834, + "step": 7340, + "valid_targets_mean": 28855.7, + "valid_targets_min": 19780 + }, + { + "epoch": 7.838847385272145, + "grad_norm": 0.13422349699284086, + "learning_rate": 2.5042801019785624e-07, + "loss": 0.398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3954845368862152, + "step": 7345, + "valid_targets_mean": 28988.1, + "valid_targets_min": 22849 + }, + { + "epoch": 7.8441835645677696, + "grad_norm": 0.12139430060409888, + "learning_rate": 2.342297913212943e-07, + "loss": 0.3957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39191263914108276, + "step": 7350, + "valid_targets_mean": 28993.3, + "valid_targets_min": 23462 + }, + { + "epoch": 7.849519743863394, + "grad_norm": 0.12127261635607517, + "learning_rate": 2.1857248672100262e-07, + "loss": 0.395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39600613713264465, + "step": 7355, + "valid_targets_mean": 28810.9, + "valid_targets_min": 22131 + }, + { + "epoch": 7.854855923159018, + "grad_norm": 0.11960436598459143, + "learning_rate": 2.0345618128840838e-07, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39346587657928467, + "step": 7360, + "valid_targets_mean": 28869.0, + "valid_targets_min": 21748 + }, + { + "epoch": 7.860192102454643, + "grad_norm": 0.11716437914975067, + "learning_rate": 1.888809569817518e-07, + "loss": 0.3976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39495378732681274, + "step": 7365, + "valid_targets_mean": 28888.0, + "valid_targets_min": 22788 + }, + { + "epoch": 7.865528281750267, + "grad_norm": 0.12118563526720992, + "learning_rate": 1.7484689282558643e-07, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39744696021080017, + "step": 7370, + "valid_targets_mean": 28844.5, + "valid_targets_min": 20332 + }, + { + "epoch": 7.870864461045891, + "grad_norm": 0.1203814795413713, + "learning_rate": 1.6135406491041282e-07, + "loss": 0.3973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3969980776309967, + "step": 7375, + "valid_targets_mean": 28854.4, + "valid_targets_min": 20845 + }, + { + "epoch": 7.876200640341516, + "grad_norm": 0.11945569838464438, + "learning_rate": 1.484025463921901e-07, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3918302059173584, + "step": 7380, + "valid_targets_mean": 28731.7, + "valid_targets_min": 18884 + }, + { + "epoch": 7.88153681963714, + "grad_norm": 0.11756182206253876, + "learning_rate": 1.359924074920249e-07, + "loss": 0.3991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40109801292419434, + "step": 7385, + "valid_targets_mean": 28831.0, + "valid_targets_min": 19023 + }, + { + "epoch": 7.886872998932764, + "grad_norm": 0.1340519609225345, + "learning_rate": 1.2412371549573864e-07, + "loss": 0.3992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39418697357177734, + "step": 7390, + "valid_targets_mean": 28842.2, + "valid_targets_min": 21875 + }, + { + "epoch": 7.892209178228389, + "grad_norm": 0.11587335971582481, + "learning_rate": 1.1279653475350094e-07, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39782729744911194, + "step": 7395, + "valid_targets_mean": 28892.1, + "valid_targets_min": 19641 + }, + { + "epoch": 7.897545357524013, + "grad_norm": 0.11750288132031445, + "learning_rate": 1.0201092667950773e-07, + "loss": 0.3975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3983398675918579, + "step": 7400, + "valid_targets_mean": 28828.2, + "valid_targets_min": 21791 + }, + { + "epoch": 7.9028815368196375, + "grad_norm": 0.13435898758171472, + "learning_rate": 9.176694975161493e-08, + "loss": 0.3983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39778202772140503, + "step": 7405, + "valid_targets_mean": 29013.8, + "valid_targets_min": 20210 + }, + { + "epoch": 7.908217716115262, + "grad_norm": 0.12001733613834027, + "learning_rate": 8.206465951104969e-08, + "loss": 0.3958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.392083078622818, + "step": 7410, + "valid_targets_mean": 28912.8, + "valid_targets_min": 20190 + }, + { + "epoch": 7.913553895410886, + "grad_norm": 0.11925283075717091, + "learning_rate": 7.290410856211071e-08, + "loss": 0.3958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3957703411579132, + "step": 7415, + "valid_targets_mean": 28886.9, + "valid_targets_min": 22025 + }, + { + "epoch": 7.9188900747065105, + "grad_norm": 0.14003332102035307, + "learning_rate": 6.428534657185736e-08, + "loss": 0.397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39609062671661377, + "step": 7420, + "valid_targets_mean": 28744.4, + "valid_targets_min": 19828 + }, + { + "epoch": 7.924226254002134, + "grad_norm": 0.11718337201752242, + "learning_rate": 5.620842026985429e-08, + "loss": 0.3954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3998251259326935, + "step": 7425, + "valid_targets_mean": 28700.6, + "valid_targets_min": 16080 + }, + { + "epoch": 7.929562433297759, + "grad_norm": 0.1408268207394538, + "learning_rate": 4.8673373447916116e-08, + "loss": 0.3979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39620524644851685, + "step": 7430, + "valid_targets_mean": 28704.3, + "valid_targets_min": 21952 + }, + { + "epoch": 7.934898612593383, + "grad_norm": 0.12045696435740527, + "learning_rate": 4.1680246959896473e-08, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3986198604106903, + "step": 7435, + "valid_targets_mean": 28944.1, + "valid_targets_min": 18722 + }, + { + "epoch": 7.940234791889008, + "grad_norm": 0.12033025936809216, + "learning_rate": 3.522907872141046e-08, + "loss": 0.3968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3966403603553772, + "step": 7440, + "valid_targets_mean": 28832.9, + "valid_targets_min": 18828 + }, + { + "epoch": 7.945570971184631, + "grad_norm": 0.1386418674094808, + "learning_rate": 2.9319903709679186e-08, + "loss": 0.3956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.396666944026947, + "step": 7445, + "valid_targets_mean": 28944.9, + "valid_targets_min": 21863 + }, + { + "epoch": 7.950907150480256, + "grad_norm": 0.11831593016496478, + "learning_rate": 2.3952753963318865e-08, + "loss": 0.3956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3972724974155426, + "step": 7450, + "valid_targets_mean": 29041.1, + "valid_targets_min": 20975 + }, + { + "epoch": 7.95624332977588, + "grad_norm": 0.11586980929359303, + "learning_rate": 1.9127658582163145e-08, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3945784270763397, + "step": 7455, + "valid_targets_mean": 28976.9, + "valid_targets_min": 18373 + }, + { + "epoch": 7.9615795090715045, + "grad_norm": 0.12002426358472632, + "learning_rate": 1.4844643727129903e-08, + "loss": 0.3969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.394758939743042, + "step": 7460, + "valid_targets_mean": 28830.2, + "valid_targets_min": 21402 + }, + { + "epoch": 7.966915688367129, + "grad_norm": 0.11820445457921111, + "learning_rate": 1.1103732620043606e-08, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3968362808227539, + "step": 7465, + "valid_targets_mean": 28914.0, + "valid_targets_min": 19185 + }, + { + "epoch": 7.972251867662753, + "grad_norm": 0.12162588336594275, + "learning_rate": 7.90494554353538e-09, + "loss": 0.3963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3936949670314789, + "step": 7470, + "valid_targets_mean": 28676.7, + "valid_targets_min": 18647 + }, + { + "epoch": 7.9775880469583775, + "grad_norm": 0.11925013033412035, + "learning_rate": 5.248299840920901e-09, + "loss": 0.3993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3953138589859009, + "step": 7475, + "valid_targets_mean": 28855.8, + "valid_targets_min": 18558 + }, + { + "epoch": 7.982924226254002, + "grad_norm": 0.12052536553549298, + "learning_rate": 3.1338099161226653e-09, + "loss": 0.3982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40008604526519775, + "step": 7480, + "valid_targets_mean": 28778.8, + "valid_targets_min": 19474 + }, + { + "epoch": 7.988260405549626, + "grad_norm": 0.11952974486435448, + "learning_rate": 1.5614872335589781e-09, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40297284722328186, + "step": 7485, + "valid_targets_mean": 28927.8, + "valid_targets_min": 22890 + }, + { + "epoch": 7.993596584845251, + "grad_norm": 0.12460476994872255, + "learning_rate": 5.313403181328447e-10, + "loss": 0.3989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3980481028556824, + "step": 7490, + "valid_targets_mean": 28616.4, + "valid_targets_min": 19327 + }, + { + "epoch": 7.998932764140875, + "grad_norm": 0.11985334396945319, + "learning_rate": 4.3374755120950685e-11, + "loss": 0.3961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3979700207710266, + "step": 7495, + "valid_targets_mean": 28892.5, + "valid_targets_min": 21667 + }, + { + "epoch": 8.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4003600478172302, + "step": 7496, + "total_flos": 4.678153526260531e+16, + "train_loss": 0.13369933033527026, + "train_runtime": 19962.4709, + "train_samples_per_second": 384.408, + "train_steps_per_second": 0.376, + "valid_targets_mean": 28802.2, + "valid_targets_min": 22042 + } + ], + "logging_steps": 5, + "max_steps": 7496, + "num_input_tokens_seen": 0, + "num_train_epochs": 8, + "save_steps": 100, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 4.678153526260531e+16, + "train_batch_size": 4, + "trial_name": null, + "trial_params": null +} diff --git a/training_args.bin b/training_args.bin new file mode 100644 index 0000000..413096e --- /dev/null +++ b/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15f7aaf72ad2c1cf0e38f961e470003353644389261a0d970975d61659e4d725 +size 8913 diff --git a/training_loss.png b/training_loss.png new file mode 100644 index 0000000..c0ca538 Binary files /dev/null and b/training_loss.png differ diff --git a/vocab.json b/vocab.json new file mode 100644 index 0000000..6c49fc6 --- /dev/null +++ b/vocab.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910 +size 2776833