commit 9760ecfab560cbc2a4a89d70b370071bf3c6dbb7 Author: ModelHub XC Date: Thu Jul 30 02:49:18 2026 +0800 初始化项目，由ModelHub XC社区提供模型 Model: Neelectric/Llama-3.1-8B-Instruct_SFT_sciencev00.12 Source: Original Platform diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..52373fe --- /dev/null +++ b/.gitattributes @@ -0,0 +1,36 @@ +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text +*.bin filter=lfs diff=lfs merge=lfs -text +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ckpt 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 +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz 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 +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl 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 +*.safetensors filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* 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 +*.wasm filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip 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 diff --git a/README.md b/README.md new file mode 100644 index 0000000..6a2bce8 --- /dev/null +++ b/README.md @@ -0,0 +1,59 @@ +--- +base_model: meta-llama/Llama-3.1-8B-Instruct +datasets: Neelectric/wildguardmix_reasoning_Llama-3.1-8B-Instruct_4096toks_refusals_only +library_name: transformers +model_name: Llama-3.1-8B-Instruct_SFT_sciencev00.12 +tags: +- generated_from_trainer +- open-r1 +- trl +- sft +licence: license +--- + +# Model Card for Llama-3.1-8B-Instruct_SFT_sciencev00.12 + +This model is a fine-tuned version of [meta-llama/Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct) on the [Neelectric/wildguardmix_reasoning_Llama-3.1-8B-Instruct_4096toks_refusals_only](https://huggingface.co/datasets/Neelectric/wildguardmix_reasoning_Llama-3.1-8B-Instruct_4096toks_refusals_only) dataset. +It has been trained using [TRL](https://github.com/huggingface/trl). + +## Quick start + +```python +from transformers import pipeline + +question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?" +generator = pipeline("text-generation", model="Neelectric/Llama-3.1-8B-Instruct_SFT_sciencev00.12", device="cuda") +output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0] +print(output["generated_text"]) +``` + +## Training procedure + +[Visualize in Weights & Biases](https://wandb.ai/neelectric/open-r1_science/runs/vhhwc1pn) + + +This model was trained with SFT. + +### Framework versions + +- TRL: 0.28.0.dev0 +- Transformers: 4.57.6 +- Pytorch: 2.9.0 +- Datasets: 4.5.0 +- Tokenizers: 0.22.2 + +## Citations + + + +Cite TRL as: + +```bibtex +@software{vonwerra2020trl, + title = {{TRL: Transformers Reinforcement Learning}}, + author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin}, + license = {Apache-2.0}, + url = {https://github.com/huggingface/trl}, + year = {2020} +} +``` \ No newline at end of file diff --git a/all_results.json b/all_results.json new file mode 100644 index 0000000..b3a3c17 --- /dev/null +++ b/all_results.json @@ -0,0 +1,8 @@ +{ + "total_flos": 6.401115957451817e+17, + "train_loss": 0.3628057053238609, + "train_runtime": 1540.8029, + "train_samples": 15696, + "train_samples_per_second": 20.374, + "train_steps_per_second": 1.273 +} \ No newline at end of file diff --git a/chat_template.jinja b/chat_template.jinja new file mode 100644 index 0000000..0ab931a --- /dev/null +++ b/chat_template.jinja @@ -0,0 +1,121 @@ +{{- bos_token }} +{%- if custom_tools is defined %} + {%- set tools = custom_tools %} +{%- endif %} +{%- if not tools_in_user_message is defined %} + {%- set tools_in_user_message = true %} +{%- endif %} +{%- if not date_string is defined %} + {%- set date_string = "26 Jul 2024" %} +{%- endif %} +{%- if not tools is defined %} + {%- set tools = none %} +{%- endif %} + +{#- This block extracts the system message, so we can slot it into the right place. #} +{%- if messages[0]['role'] == 'system' %} + {%- set system_message = messages[0]['content']|trim %} + {%- set messages = messages[1:] %} +{%- else %} + {%- set system_message = "You are a helpful AI Assistant that provides well-reasoned and detailed responses. You first think about the reasoning process as an internal monologue and then provide the user with the answer. Respond in the following format: +... + + +... +" %} +{%- endif %} + +{#- System message + builtin tools #} +{{- "<|start_header_id|>system<|end_header_id|>\n\n" }} +{%- if builtin_tools is defined or tools is not none %} + {{- "Environment: ipython\n" }} +{%- endif %} +{%- if builtin_tools is defined %} + {{- "Tools: " + builtin_tools | reject('equalto', 'code_interpreter') | join(", ") + "\n\n"}} +{%- endif %} +{{- "Cutting Knowledge Date: December 2023\n" }} +{{- "Today Date: " + date_string + "\n\n" }} +{%- if tools is not none and not tools_in_user_message %} + {{- "You have access to the following functions. To call a function, please respond with JSON for a function call." }} + {{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }} + {{- "Do not use variables.\n\n" }} + {%- for t in tools %} + {{- t | tojson(indent=4) }} + {{- "\n\n" }} + {%- endfor %} +{%- endif %} +{{- system_message }} +{{- "<|eot_id|>" }} + +{#- Custom tools are passed in a user message with some extra guidance #} +{%- if tools_in_user_message and not tools is none %} + {#- Extract the first user message so we can plug it in here #} + {%- if messages | length != 0 %} + {%- set first_user_message = messages[0]['content']|trim %} + {%- set messages = messages[1:] %} + {%- else %} + {{- raise_exception("Cannot put tools in the first user message when there's no first user message!") }} + {%- endif %} + {{- '<|start_header_id|>user<|end_header_id|>\n\n' -}} + {{- "Given the following functions, please respond with a JSON for a function call " }} + {{- "with its proper arguments that best answers the given prompt.\n\n" }} + {{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }} + {{- "Do not use variables.\n\n" }} + {%- for t in tools %} + {{- t | tojson(indent=4) }} + {{- "\n\n" }} + {%- endfor %} + {{- first_user_message + "<|eot_id|>"}} +{%- endif %} + +{%- for message in messages %} + {%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %} + {%- if message['role'] == 'assistant' %} + {{- '<|start_header_id|>assistant<|end_header_id|>\n\n' }} + {% generation %} + {{- message['content'] | trim + '<|eot_id|>' }} + {% endgeneration %} + {%- else %} + {{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' }} + {%- endif %} + {%- elif 'tool_calls' in message %} + {%- if not message.tool_calls|length == 1 %} + {{- raise_exception("This model only supports single tool-calls at once!") }} + {%- endif %} + {%- set tool_call = message.tool_calls[0].function %} + {{- '<|start_header_id|>assistant<|end_header_id|>\n\n' -}} + {% generation %} + {%- if builtin_tools is defined and tool_call.name in builtin_tools %} + {{- "<|python_tag|>" + tool_call.name + ".call(" }} + {%- for arg_name, arg_val in tool_call.arguments | items %} + {{- arg_name + '="' + arg_val + '"' }} + {%- if not loop.last %} + {{- ", " }} + {%- endif %} + {%- endfor %} + {{- ")" }} + {%- else %} + {{- '{"name": "' + tool_call.name + '", ' }} + {{- '"parameters": ' }} + {{- tool_call.arguments | tojson }} + {{- "}" }} + {%- endif %} + {%- if builtin_tools is defined %} + {{- "<|eom_id|>" }} + {%- else %} + {{- "<|eot_id|>" }} + {%- endif %} + {% endgeneration %} + {%- elif message.role == "tool" or message.role == "ipython" %} + {{- "<|start_header_id|>ipython<|end_header_id|>\n\n" }} + {%- if message.content is mapping or message.content is iterable %} + {{- message.content | tojson }} + {%- else %} + {{- message.content }} + {%- endif %} + {{- "<|eot_id|>" }} + {%- endif %} +{%- endfor %} +{%- if add_generation_prompt %} + {{- '<|start_header_id|>assistant<|end_header_id|>\n\n' }} +{%- endif %} \ No newline at end of file diff --git a/config.json b/config.json new file mode 100644 index 0000000..e1d9068 --- /dev/null +++ b/config.json @@ -0,0 +1,35 @@ +{ + "architectures": [ + "LlamaForCausalLM" + ], + "attention_bias": false, + "attention_dropout": 0.0, + "bos_token_id": 128000, + "dtype": "bfloat16", + "eos_token_id": 128009, + "head_dim": 128, + "hidden_act": "silu", + "hidden_size": 4096, + "initializer_range": 0.02, + "intermediate_size": 14336, + "max_position_embeddings": 131072, + "mlp_bias": false, + "model_type": "llama", + "num_attention_heads": 32, + "num_hidden_layers": 32, + "num_key_value_heads": 8, + "pretraining_tp": 1, + "rms_norm_eps": 1e-05, + "rope_scaling": { + "factor": 8.0, + "high_freq_factor": 4.0, + "low_freq_factor": 1.0, + "original_max_position_embeddings": 8192, + "rope_type": "llama3" + }, + "rope_theta": 500000.0, + "tie_word_embeddings": false, + "transformers_version": "4.57.6", + "use_cache": true, + "vocab_size": 128256 +} diff --git a/generation_config.json b/generation_config.json new file mode 100644 index 0000000..1996dc1 --- /dev/null +++ b/generation_config.json @@ -0,0 +1,8 @@ +{ + "bos_token_id": 128000, + "do_sample": true, + "eos_token_id": 128009, + "temperature": 0.6, + "top_p": 0.9, + "transformers_version": "4.57.6" +} diff --git a/model-00001-of-00004.safetensors b/model-00001-of-00004.safetensors new file mode 100644 index 0000000..edfe17e --- /dev/null +++ b/model-00001-of-00004.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d0d1705054f75b0d0159fdc7ae01e00e59bfbfdf8399a53208d986fd18ad5b1 +size 4976698672 diff --git a/model-00002-of-00004.safetensors b/model-00002-of-00004.safetensors new file mode 100644 index 0000000..9ac639f --- /dev/null +++ b/model-00002-of-00004.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f861ad70c3e2a807d846993cedeab6462856f407a97c42dcbdb367fed9a5a527 +size 4999802720 diff --git a/model-00003-of-00004.safetensors b/model-00003-of-00004.safetensors new file mode 100644 index 0000000..aab791c --- /dev/null +++ b/model-00003-of-00004.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62951702a57408134866944803bc70d2a33552b2e632f882797acd61de159720 +size 4915916176 diff --git a/model-00004-of-00004.safetensors b/model-00004-of-00004.safetensors new file mode 100644 index 0000000..3c4772c --- /dev/null +++ b/model-00004-of-00004.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dbe3d968d054fcb083807c2b5518fe0580352bb0fce14e53418a0fb082eda2b1 +size 1168138808 diff --git a/model.safetensors.index.json b/model.safetensors.index.json new file mode 100644 index 0000000..5c64f1e --- /dev/null +++ b/model.safetensors.index.json @@ -0,0 +1,299 @@ +{ + "metadata": { + "total_parameters": 8030261248, + "total_size": 16060522496 + }, + "weight_map": { + "lm_head.weight": "model-00004-of-00004.safetensors", + "model.embed_tokens.weight": "model-00001-of-00004.safetensors", + "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.16.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.17.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.17.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.18.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.19.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.19.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.19.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.20.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.20.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.20.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.20.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.20.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.26.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.27.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.27.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.27.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.28.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.28.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.28.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.29.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.29.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.29.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.30.input_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.30.mlp.down_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.30.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.30.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.30.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", + "model.layers.30.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.30.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.30.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.30.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.31.input_layernorm.weight": "model-00004-of-00004.safetensors", + "model.layers.31.mlp.down_proj.weight": "model-00004-of-00004.safetensors", + "model.layers.31.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.31.mlp.up_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.31.post_attention_layernorm.weight": "model-00004-of-00004.safetensors", + "model.layers.31.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.31.self_attn.o_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.31.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.31.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", + "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.input_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.8.mlp.down_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.mlp.gate_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.mlp.up_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00004.safetensors", + "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00004.safetensors", + "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors", + "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors", + "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors", + "model.norm.weight": "model-00004-of-00004.safetensors" + } +} diff --git a/special_tokens_map.json b/special_tokens_map.json new file mode 100644 index 0000000..e8f05fa --- /dev/null +++ b/special_tokens_map.json @@ -0,0 +1,10 @@ +{ + "bos_token": { + "content": "<|begin_of_text|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false + }, + "eos_token": "<|eot_id|>" +} diff --git a/tokenizer.json b/tokenizer.json new file mode 100644 index 0000000..1c1d8d5 --- /dev/null +++ b/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b9e4e7fb171f92fd137b777cc2714bf87d11576700a1dcd7a399e7bbe39537b +size 17209920 diff --git a/tokenizer_config.json b/tokenizer_config.json new file mode 100644 index 0000000..8b0c7c1 --- /dev/null +++ b/tokenizer_config.json @@ -0,0 +1,2062 @@ +{ + "added_tokens_decoder": { + "128000": { + "content": "<|begin_of_text|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128001": { + "content": "<|end_of_text|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128002": { + "content": "<|reserved_special_token_0|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128003": { + "content": "<|reserved_special_token_1|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128004": { + "content": "<|finetune_right_pad_id|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128005": { + "content": "<|reserved_special_token_2|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128006": { + "content": "<|start_header_id|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128007": { + "content": "<|end_header_id|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128008": { + "content": "<|eom_id|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128009": { + "content": "<|eot_id|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128010": { + "content": "<|python_tag|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128011": { + "content": "<|reserved_special_token_3|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128012": { + "content": "<|reserved_special_token_4|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128013": { + "content": "<|reserved_special_token_5|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128014": { + "content": "<|reserved_special_token_6|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128015": { + "content": "<|reserved_special_token_7|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128016": { + "content": "<|reserved_special_token_8|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128017": { + "content": "<|reserved_special_token_9|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128018": { + "content": "<|reserved_special_token_10|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128019": { + "content": "<|reserved_special_token_11|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128020": { + "content": "<|reserved_special_token_12|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128021": { + "content": "<|reserved_special_token_13|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128022": { + "content": "<|reserved_special_token_14|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128023": { + "content": "<|reserved_special_token_15|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128024": { + "content": "<|reserved_special_token_16|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128025": { + "content": "<|reserved_special_token_17|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128026": { + "content": "<|reserved_special_token_18|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128027": { + "content": "<|reserved_special_token_19|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128028": { + "content": "<|reserved_special_token_20|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128029": { + "content": "<|reserved_special_token_21|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128030": { + "content": "<|reserved_special_token_22|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128031": { + "content": "<|reserved_special_token_23|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128032": { + "content": "<|reserved_special_token_24|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128033": { + "content": "<|reserved_special_token_25|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128034": { + "content": "<|reserved_special_token_26|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128035": { + "content": "<|reserved_special_token_27|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128036": { + "content": "<|reserved_special_token_28|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128037": { + "content": "<|reserved_special_token_29|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128038": { + "content": "<|reserved_special_token_30|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128039": { + "content": "<|reserved_special_token_31|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128040": { + "content": "<|reserved_special_token_32|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128041": { + "content": "<|reserved_special_token_33|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128042": { + "content": "<|reserved_special_token_34|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128043": { + "content": "<|reserved_special_token_35|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128044": { + "content": "<|reserved_special_token_36|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128045": { + "content": "<|reserved_special_token_37|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128046": { + "content": "<|reserved_special_token_38|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128047": { + "content": "<|reserved_special_token_39|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128048": { + "content": "<|reserved_special_token_40|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128049": { + "content": "<|reserved_special_token_41|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128050": { + "content": "<|reserved_special_token_42|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128051": { + "content": "<|reserved_special_token_43|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128052": { + "content": "<|reserved_special_token_44|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128053": { + "content": "<|reserved_special_token_45|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128054": { + "content": "<|reserved_special_token_46|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128055": { + "content": "<|reserved_special_token_47|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128056": { + "content": "<|reserved_special_token_48|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128057": { + "content": "<|reserved_special_token_49|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128058": { + "content": "<|reserved_special_token_50|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128059": { + "content": "<|reserved_special_token_51|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128060": { + "content": "<|reserved_special_token_52|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128061": { + "content": "<|reserved_special_token_53|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128062": { + "content": "<|reserved_special_token_54|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128063": { + "content": "<|reserved_special_token_55|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128064": { + "content": "<|reserved_special_token_56|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128065": { + "content": "<|reserved_special_token_57|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128066": { + "content": "<|reserved_special_token_58|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128067": { + "content": "<|reserved_special_token_59|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128068": { + "content": "<|reserved_special_token_60|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128069": { + "content": "<|reserved_special_token_61|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128070": { + "content": "<|reserved_special_token_62|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128071": { + "content": "<|reserved_special_token_63|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128072": { + "content": "<|reserved_special_token_64|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128073": { + "content": "<|reserved_special_token_65|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128074": { + "content": "<|reserved_special_token_66|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128075": { + "content": "<|reserved_special_token_67|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128076": { + "content": "<|reserved_special_token_68|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128077": { + "content": "<|reserved_special_token_69|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128078": { + "content": "<|reserved_special_token_70|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128079": { + "content": "<|reserved_special_token_71|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128080": { + "content": "<|reserved_special_token_72|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128081": { + "content": "<|reserved_special_token_73|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128082": { + "content": "<|reserved_special_token_74|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128083": { + "content": "<|reserved_special_token_75|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128084": { + "content": "<|reserved_special_token_76|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128085": { + "content": "<|reserved_special_token_77|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128086": { + "content": "<|reserved_special_token_78|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128087": { + "content": "<|reserved_special_token_79|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128088": { + "content": "<|reserved_special_token_80|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128089": { + "content": "<|reserved_special_token_81|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128090": { + "content": "<|reserved_special_token_82|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128091": { + "content": "<|reserved_special_token_83|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128092": { + "content": "<|reserved_special_token_84|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128093": { + "content": "<|reserved_special_token_85|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128094": { + "content": "<|reserved_special_token_86|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128095": { + "content": "<|reserved_special_token_87|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128096": { + "content": "<|reserved_special_token_88|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128097": { + "content": "<|reserved_special_token_89|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128098": { + "content": "<|reserved_special_token_90|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128099": { + "content": "<|reserved_special_token_91|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128100": { + "content": "<|reserved_special_token_92|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128101": { + "content": "<|reserved_special_token_93|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128102": { + "content": "<|reserved_special_token_94|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128103": { + "content": "<|reserved_special_token_95|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128104": { + "content": "<|reserved_special_token_96|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128105": { + "content": "<|reserved_special_token_97|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128106": { + "content": "<|reserved_special_token_98|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128107": { + "content": "<|reserved_special_token_99|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128108": { + "content": "<|reserved_special_token_100|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128109": { + "content": "<|reserved_special_token_101|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128110": { + "content": "<|reserved_special_token_102|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128111": { + "content": "<|reserved_special_token_103|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128112": { + "content": "<|reserved_special_token_104|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128113": { + "content": "<|reserved_special_token_105|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128114": { + "content": "<|reserved_special_token_106|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128115": { + "content": "<|reserved_special_token_107|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128116": { + "content": "<|reserved_special_token_108|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128117": { + "content": "<|reserved_special_token_109|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128118": { + "content": "<|reserved_special_token_110|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128119": { + "content": "<|reserved_special_token_111|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128120": { + "content": "<|reserved_special_token_112|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128121": { + "content": "<|reserved_special_token_113|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128122": { + "content": "<|reserved_special_token_114|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128123": { + "content": "<|reserved_special_token_115|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128124": { + "content": "<|reserved_special_token_116|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128125": { + "content": "<|reserved_special_token_117|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128126": { + "content": "<|reserved_special_token_118|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128127": { + "content": "<|reserved_special_token_119|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128128": { + "content": "<|reserved_special_token_120|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128129": { + "content": "<|reserved_special_token_121|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128130": { + "content": "<|reserved_special_token_122|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128131": { + "content": "<|reserved_special_token_123|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128132": { + "content": "<|reserved_special_token_124|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128133": { + "content": "<|reserved_special_token_125|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128134": { + "content": "<|reserved_special_token_126|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128135": { + "content": "<|reserved_special_token_127|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128136": { + "content": "<|reserved_special_token_128|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128137": { + "content": "<|reserved_special_token_129|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128138": { + "content": "<|reserved_special_token_130|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128139": { + "content": "<|reserved_special_token_131|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128140": { + "content": "<|reserved_special_token_132|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128141": { + "content": "<|reserved_special_token_133|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128142": { + "content": "<|reserved_special_token_134|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128143": { + "content": "<|reserved_special_token_135|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128144": { + "content": "<|reserved_special_token_136|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128145": { + "content": "<|reserved_special_token_137|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128146": { + "content": "<|reserved_special_token_138|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128147": { + "content": "<|reserved_special_token_139|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128148": { + "content": "<|reserved_special_token_140|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128149": { + "content": "<|reserved_special_token_141|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128150": { + "content": "<|reserved_special_token_142|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128151": { + "content": "<|reserved_special_token_143|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128152": { + "content": "<|reserved_special_token_144|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128153": { + "content": "<|reserved_special_token_145|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128154": { + "content": "<|reserved_special_token_146|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128155": { + "content": "<|reserved_special_token_147|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128156": { + "content": "<|reserved_special_token_148|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128157": { + "content": "<|reserved_special_token_149|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128158": { + "content": "<|reserved_special_token_150|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128159": { + "content": "<|reserved_special_token_151|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128160": { + "content": "<|reserved_special_token_152|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128161": { + "content": "<|reserved_special_token_153|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128162": { + "content": "<|reserved_special_token_154|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128163": { + "content": "<|reserved_special_token_155|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128164": { + "content": "<|reserved_special_token_156|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128165": { + "content": "<|reserved_special_token_157|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128166": { + "content": "<|reserved_special_token_158|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128167": { + "content": "<|reserved_special_token_159|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128168": { + "content": "<|reserved_special_token_160|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128169": { + "content": "<|reserved_special_token_161|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128170": { + "content": "<|reserved_special_token_162|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128171": { + "content": "<|reserved_special_token_163|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128172": { + "content": "<|reserved_special_token_164|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128173": { + "content": "<|reserved_special_token_165|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128174": { + "content": "<|reserved_special_token_166|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128175": { + "content": "<|reserved_special_token_167|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128176": { + "content": "<|reserved_special_token_168|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128177": { + "content": "<|reserved_special_token_169|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128178": { + "content": "<|reserved_special_token_170|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128179": { + "content": "<|reserved_special_token_171|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128180": { + "content": "<|reserved_special_token_172|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128181": { + "content": "<|reserved_special_token_173|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128182": { + "content": "<|reserved_special_token_174|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128183": { + "content": "<|reserved_special_token_175|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128184": { + "content": "<|reserved_special_token_176|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128185": { + "content": "<|reserved_special_token_177|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128186": { + "content": "<|reserved_special_token_178|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128187": { + "content": "<|reserved_special_token_179|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128188": { + "content": "<|reserved_special_token_180|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128189": { + "content": "<|reserved_special_token_181|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128190": { + "content": "<|reserved_special_token_182|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128191": { + "content": "<|reserved_special_token_183|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128192": { + "content": "<|reserved_special_token_184|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128193": { + "content": "<|reserved_special_token_185|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128194": { + "content": "<|reserved_special_token_186|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128195": { + "content": "<|reserved_special_token_187|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128196": { + "content": "<|reserved_special_token_188|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128197": { + "content": "<|reserved_special_token_189|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128198": { + "content": "<|reserved_special_token_190|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128199": { + "content": "<|reserved_special_token_191|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128200": { + "content": "<|reserved_special_token_192|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128201": { + "content": "<|reserved_special_token_193|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128202": { + "content": "<|reserved_special_token_194|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128203": { + "content": "<|reserved_special_token_195|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128204": { + "content": "<|reserved_special_token_196|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128205": { + "content": "<|reserved_special_token_197|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128206": { + "content": "<|reserved_special_token_198|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128207": { + "content": "<|reserved_special_token_199|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128208": { + "content": "<|reserved_special_token_200|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128209": { + "content": "<|reserved_special_token_201|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128210": { + "content": "<|reserved_special_token_202|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128211": { + "content": "<|reserved_special_token_203|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128212": { + "content": "<|reserved_special_token_204|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128213": { + "content": "<|reserved_special_token_205|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128214": { + "content": "<|reserved_special_token_206|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128215": { + "content": "<|reserved_special_token_207|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128216": { + "content": "<|reserved_special_token_208|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128217": { + "content": "<|reserved_special_token_209|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128218": { + "content": "<|reserved_special_token_210|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128219": { + "content": "<|reserved_special_token_211|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128220": { + "content": "<|reserved_special_token_212|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128221": { + "content": "<|reserved_special_token_213|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128222": { + "content": "<|reserved_special_token_214|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128223": { + "content": "<|reserved_special_token_215|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128224": { + "content": "<|reserved_special_token_216|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128225": { + "content": "<|reserved_special_token_217|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128226": { + "content": "<|reserved_special_token_218|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128227": { + "content": "<|reserved_special_token_219|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128228": { + "content": "<|reserved_special_token_220|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128229": { + "content": "<|reserved_special_token_221|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128230": { + "content": "<|reserved_special_token_222|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128231": { + "content": "<|reserved_special_token_223|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128232": { + "content": "<|reserved_special_token_224|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128233": { + "content": "<|reserved_special_token_225|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128234": { + "content": "<|reserved_special_token_226|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128235": { + "content": "<|reserved_special_token_227|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128236": { + "content": "<|reserved_special_token_228|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128237": { + "content": "<|reserved_special_token_229|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128238": { + "content": "<|reserved_special_token_230|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128239": { + "content": "<|reserved_special_token_231|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128240": { + "content": "<|reserved_special_token_232|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128241": { + "content": "<|reserved_special_token_233|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128242": { + "content": "<|reserved_special_token_234|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128243": { + "content": "<|reserved_special_token_235|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128244": { + "content": "<|reserved_special_token_236|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128245": { + "content": "<|reserved_special_token_237|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128246": { + "content": "<|reserved_special_token_238|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128247": { + "content": "<|reserved_special_token_239|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128248": { + "content": "<|reserved_special_token_240|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128249": { + "content": "<|reserved_special_token_241|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128250": { + "content": "<|reserved_special_token_242|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128251": { + "content": "<|reserved_special_token_243|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128252": { + "content": "<|reserved_special_token_244|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128253": { + "content": "<|reserved_special_token_245|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128254": { + "content": "<|reserved_special_token_246|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "128255": { + "content": "<|reserved_special_token_247|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + } + }, + "bos_token": "<|begin_of_text|>", + "clean_up_tokenization_spaces": true, + "eos_token": "<|eot_id|>", + "extra_special_tokens": {}, + "model_input_names": [ + "input_ids", + "attention_mask" + ], + "model_max_length": 131072, + "tokenizer_class": "PreTrainedTokenizerFast" +} diff --git a/train_results.json b/train_results.json new file mode 100644 index 0000000..b3a3c17 --- /dev/null +++ b/train_results.json @@ -0,0 +1,8 @@ +{ + "total_flos": 6.401115957451817e+17, + "train_loss": 0.3628057053238609, + "train_runtime": 1540.8029, + "train_samples": 15696, + "train_samples_per_second": 20.374, + "train_steps_per_second": 1.273 +} \ No newline at end of file diff --git a/trainer_state.json b/trainer_state.json new file mode 100644 index 0000000..94c9e7f --- /dev/null +++ b/trainer_state.json @@ -0,0 +1,17701 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 2.0, + "eval_steps": 500, + "global_step": 1962, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0010193679918450561, + "grad_norm": 14.944640159606934, + "learning_rate": 0.0, + "loss": 0.6867, + "mean_token_accuracy": 0.8096808791160583, + "num_tokens": 7262.0, + "step": 1 + }, + { + "epoch": 0.0020387359836901123, + "grad_norm": 16.410602569580078, + "learning_rate": 5.076142131979695e-09, + "loss": 0.7043, + "mean_token_accuracy": 0.8117748498916626, + "num_tokens": 14306.0, + "step": 2 + }, + { + "epoch": 0.0030581039755351682, + "grad_norm": 18.264877319335938, + "learning_rate": 1.015228426395939e-08, + "loss": 0.734, + "mean_token_accuracy": 0.8022207617759705, + "num_tokens": 21029.0, + "step": 3 + }, + { + "epoch": 0.004077471967380225, + "grad_norm": 20.244123458862305, + "learning_rate": 1.5228426395939085e-08, + "loss": 0.7414, + "mean_token_accuracy": 0.8089109659194946, + "num_tokens": 27115.0, + "step": 4 + }, + { + "epoch": 0.0050968399592252805, + "grad_norm": 19.42837905883789, + "learning_rate": 2.030456852791878e-08, + "loss": 0.7241, + "mean_token_accuracy": 0.8209551572799683, + "num_tokens": 33179.0, + "step": 5 + }, + { + "epoch": 0.0061162079510703364, + "grad_norm": 18.210630416870117, + "learning_rate": 2.5380710659898475e-08, + "loss": 0.7419, + "mean_token_accuracy": 0.8005608320236206, + "num_tokens": 40268.0, + "step": 6 + }, + { + "epoch": 0.007135575942915392, + "grad_norm": 15.372761726379395, + "learning_rate": 3.045685279187817e-08, + "loss": 0.6493, + "mean_token_accuracy": 0.8156945705413818, + "num_tokens": 48055.0, + "step": 7 + }, + { + "epoch": 0.00815494393476045, + "grad_norm": 16.858036041259766, + "learning_rate": 3.553299492385787e-08, + "loss": 0.687, + "mean_token_accuracy": 0.8031160831451416, + "num_tokens": 54674.0, + "step": 8 + }, + { + "epoch": 0.009174311926605505, + "grad_norm": 15.499384880065918, + "learning_rate": 4.060913705583756e-08, + "loss": 0.6036, + "mean_token_accuracy": 0.8212350606918335, + "num_tokens": 62585.0, + "step": 9 + }, + { + "epoch": 0.010193679918450561, + "grad_norm": 14.529195785522461, + "learning_rate": 4.5685279187817256e-08, + "loss": 0.6491, + "mean_token_accuracy": 0.823378324508667, + "num_tokens": 70918.0, + "step": 10 + }, + { + "epoch": 0.011213047910295617, + "grad_norm": 13.455702781677246, + "learning_rate": 5.076142131979695e-08, + "loss": 0.6394, + "mean_token_accuracy": 0.8209995031356812, + "num_tokens": 80504.0, + "step": 11 + }, + { + "epoch": 0.012232415902140673, + "grad_norm": 17.388341903686523, + "learning_rate": 5.5837563451776646e-08, + "loss": 0.7193, + "mean_token_accuracy": 0.8050376176834106, + "num_tokens": 87302.0, + "step": 12 + }, + { + "epoch": 0.013251783893985729, + "grad_norm": 21.264793395996094, + "learning_rate": 6.091370558375634e-08, + "loss": 0.8116, + "mean_token_accuracy": 0.7946767210960388, + "num_tokens": 93417.0, + "step": 13 + }, + { + "epoch": 0.014271151885830785, + "grad_norm": 14.617177963256836, + "learning_rate": 6.598984771573604e-08, + "loss": 0.698, + "mean_token_accuracy": 0.8014162182807922, + "num_tokens": 101559.0, + "step": 14 + }, + { + "epoch": 0.01529051987767584, + "grad_norm": 18.83754539489746, + "learning_rate": 7.106598984771574e-08, + "loss": 0.7305, + "mean_token_accuracy": 0.8063344955444336, + "num_tokens": 108916.0, + "step": 15 + }, + { + "epoch": 0.0163098878695209, + "grad_norm": 14.805863380432129, + "learning_rate": 7.614213197969544e-08, + "loss": 0.6255, + "mean_token_accuracy": 0.82572340965271, + "num_tokens": 117124.0, + "step": 16 + }, + { + "epoch": 0.017329255861365953, + "grad_norm": 12.511829376220703, + "learning_rate": 8.121827411167512e-08, + "loss": 0.6052, + "mean_token_accuracy": 0.8271230459213257, + "num_tokens": 125396.0, + "step": 17 + }, + { + "epoch": 0.01834862385321101, + "grad_norm": 12.123272895812988, + "learning_rate": 8.629441624365482e-08, + "loss": 0.5646, + "mean_token_accuracy": 0.8366028070449829, + "num_tokens": 134886.0, + "step": 18 + }, + { + "epoch": 0.019367991845056064, + "grad_norm": 18.86951446533203, + "learning_rate": 9.137055837563451e-08, + "loss": 0.7575, + "mean_token_accuracy": 0.8056033849716187, + "num_tokens": 141139.0, + "step": 19 + }, + { + "epoch": 0.020387359836901122, + "grad_norm": 18.4528865814209, + "learning_rate": 9.64467005076142e-08, + "loss": 0.7007, + "mean_token_accuracy": 0.818597674369812, + "num_tokens": 147289.0, + "step": 20 + }, + { + "epoch": 0.021406727828746176, + "grad_norm": 16.559247970581055, + "learning_rate": 1.015228426395939e-07, + "loss": 0.6486, + "mean_token_accuracy": 0.8231261372566223, + "num_tokens": 154413.0, + "step": 21 + }, + { + "epoch": 0.022426095820591234, + "grad_norm": 26.239561080932617, + "learning_rate": 1.065989847715736e-07, + "loss": 0.825, + "mean_token_accuracy": 0.8125928640365601, + "num_tokens": 159677.0, + "step": 22 + }, + { + "epoch": 0.023445463812436288, + "grad_norm": 18.30411720275879, + "learning_rate": 1.1167512690355329e-07, + "loss": 0.7147, + "mean_token_accuracy": 0.796891450881958, + "num_tokens": 167099.0, + "step": 23 + }, + { + "epoch": 0.024464831804281346, + "grad_norm": 18.95929527282715, + "learning_rate": 1.1675126903553299e-07, + "loss": 0.7056, + "mean_token_accuracy": 0.8060421943664551, + "num_tokens": 173378.0, + "step": 24 + }, + { + "epoch": 0.0254841997961264, + "grad_norm": 14.205365180969238, + "learning_rate": 1.2182741116751268e-07, + "loss": 0.5997, + "mean_token_accuracy": 0.834286093711853, + "num_tokens": 181912.0, + "step": 25 + }, + { + "epoch": 0.026503567787971458, + "grad_norm": 19.593332290649414, + "learning_rate": 1.2690355329949238e-07, + "loss": 0.7091, + "mean_token_accuracy": 0.8165711760520935, + "num_tokens": 188237.0, + "step": 26 + }, + { + "epoch": 0.027522935779816515, + "grad_norm": 15.025238990783691, + "learning_rate": 1.3197969543147207e-07, + "loss": 0.6045, + "mean_token_accuracy": 0.8360809087753296, + "num_tokens": 195449.0, + "step": 27 + }, + { + "epoch": 0.02854230377166157, + "grad_norm": 16.308902740478516, + "learning_rate": 1.3705583756345177e-07, + "loss": 0.6881, + "mean_token_accuracy": 0.806698739528656, + "num_tokens": 203156.0, + "step": 28 + }, + { + "epoch": 0.029561671763506627, + "grad_norm": 21.182804107666016, + "learning_rate": 1.421319796954315e-07, + "loss": 0.7982, + "mean_token_accuracy": 0.7777559161186218, + "num_tokens": 208706.0, + "step": 29 + }, + { + "epoch": 0.03058103975535168, + "grad_norm": 12.200752258300781, + "learning_rate": 1.4720812182741116e-07, + "loss": 0.5866, + "mean_token_accuracy": 0.817126452922821, + "num_tokens": 217732.0, + "step": 30 + }, + { + "epoch": 0.03160040774719674, + "grad_norm": 15.084386825561523, + "learning_rate": 1.5228426395939088e-07, + "loss": 0.6665, + "mean_token_accuracy": 0.7957025766372681, + "num_tokens": 225985.0, + "step": 31 + }, + { + "epoch": 0.0326197757390418, + "grad_norm": 21.578975677490234, + "learning_rate": 1.5736040609137055e-07, + "loss": 0.7283, + "mean_token_accuracy": 0.8046397566795349, + "num_tokens": 231178.0, + "step": 32 + }, + { + "epoch": 0.03363914373088685, + "grad_norm": 17.151220321655273, + "learning_rate": 1.6243654822335024e-07, + "loss": 0.6753, + "mean_token_accuracy": 0.8152144551277161, + "num_tokens": 237070.0, + "step": 33 + }, + { + "epoch": 0.034658511722731905, + "grad_norm": 15.455815315246582, + "learning_rate": 1.6751269035532994e-07, + "loss": 0.6523, + "mean_token_accuracy": 0.8001407384872437, + "num_tokens": 244530.0, + "step": 34 + }, + { + "epoch": 0.03567787971457696, + "grad_norm": 13.894721984863281, + "learning_rate": 1.7258883248730963e-07, + "loss": 0.6182, + "mean_token_accuracy": 0.8094278573989868, + "num_tokens": 252898.0, + "step": 35 + }, + { + "epoch": 0.03669724770642202, + "grad_norm": 18.730148315429688, + "learning_rate": 1.7766497461928933e-07, + "loss": 0.6559, + "mean_token_accuracy": 0.8092378377914429, + "num_tokens": 259786.0, + "step": 36 + }, + { + "epoch": 0.03771661569826707, + "grad_norm": 15.029507637023926, + "learning_rate": 1.8274111675126902e-07, + "loss": 0.5924, + "mean_token_accuracy": 0.8279557228088379, + "num_tokens": 267465.0, + "step": 37 + }, + { + "epoch": 0.03873598369011213, + "grad_norm": 17.411714553833008, + "learning_rate": 1.8781725888324872e-07, + "loss": 0.616, + "mean_token_accuracy": 0.8239544630050659, + "num_tokens": 273599.0, + "step": 38 + }, + { + "epoch": 0.039755351681957186, + "grad_norm": 12.756753921508789, + "learning_rate": 1.928934010152284e-07, + "loss": 0.5909, + "mean_token_accuracy": 0.8141524791717529, + "num_tokens": 282553.0, + "step": 39 + }, + { + "epoch": 0.040774719673802244, + "grad_norm": 14.924580574035645, + "learning_rate": 1.9796954314720813e-07, + "loss": 0.6112, + "mean_token_accuracy": 0.810245156288147, + "num_tokens": 289445.0, + "step": 40 + }, + { + "epoch": 0.0417940876656473, + "grad_norm": 14.659953117370605, + "learning_rate": 2.030456852791878e-07, + "loss": 0.5142, + "mean_token_accuracy": 0.8301295638084412, + "num_tokens": 296579.0, + "step": 41 + }, + { + "epoch": 0.04281345565749235, + "grad_norm": 11.598182678222656, + "learning_rate": 2.0812182741116752e-07, + "loss": 0.5165, + "mean_token_accuracy": 0.8355472087860107, + "num_tokens": 304188.0, + "step": 42 + }, + { + "epoch": 0.04383282364933741, + "grad_norm": 10.887053489685059, + "learning_rate": 2.131979695431472e-07, + "loss": 0.4738, + "mean_token_accuracy": 0.8349006175994873, + "num_tokens": 311669.0, + "step": 43 + }, + { + "epoch": 0.04485219164118247, + "grad_norm": 9.62663459777832, + "learning_rate": 2.182741116751269e-07, + "loss": 0.4657, + "mean_token_accuracy": 0.8349779844284058, + "num_tokens": 319656.0, + "step": 44 + }, + { + "epoch": 0.045871559633027525, + "grad_norm": 12.988221168518066, + "learning_rate": 2.2335025380710658e-07, + "loss": 0.5577, + "mean_token_accuracy": 0.8127996325492859, + "num_tokens": 325785.0, + "step": 45 + }, + { + "epoch": 0.046890927624872576, + "grad_norm": 8.35251235961914, + "learning_rate": 2.2842639593908628e-07, + "loss": 0.44, + "mean_token_accuracy": 0.8488821983337402, + "num_tokens": 335417.0, + "step": 46 + }, + { + "epoch": 0.047910295616717634, + "grad_norm": 8.189242362976074, + "learning_rate": 2.3350253807106597e-07, + "loss": 0.4531, + "mean_token_accuracy": 0.8401865363121033, + "num_tokens": 344507.0, + "step": 47 + }, + { + "epoch": 0.04892966360856269, + "grad_norm": 9.642645835876465, + "learning_rate": 2.3857868020304564e-07, + "loss": 0.4659, + "mean_token_accuracy": 0.8355200886726379, + "num_tokens": 352146.0, + "step": 48 + }, + { + "epoch": 0.04994903160040775, + "grad_norm": 8.397224426269531, + "learning_rate": 2.4365482233502536e-07, + "loss": 0.4086, + "mean_token_accuracy": 0.8618808388710022, + "num_tokens": 359818.0, + "step": 49 + }, + { + "epoch": 0.0509683995922528, + "grad_norm": 8.776206016540527, + "learning_rate": 2.487309644670051e-07, + "loss": 0.4545, + "mean_token_accuracy": 0.834607720375061, + "num_tokens": 366709.0, + "step": 50 + }, + { + "epoch": 0.05198776758409786, + "grad_norm": 7.493223190307617, + "learning_rate": 2.5380710659898475e-07, + "loss": 0.447, + "mean_token_accuracy": 0.8415888547897339, + "num_tokens": 375845.0, + "step": 51 + }, + { + "epoch": 0.053007135575942915, + "grad_norm": 7.300378799438477, + "learning_rate": 2.588832487309645e-07, + "loss": 0.4468, + "mean_token_accuracy": 0.8342018127441406, + "num_tokens": 383789.0, + "step": 52 + }, + { + "epoch": 0.05402650356778797, + "grad_norm": 8.0009126663208, + "learning_rate": 2.6395939086294414e-07, + "loss": 0.4484, + "mean_token_accuracy": 0.8332052230834961, + "num_tokens": 390636.0, + "step": 53 + }, + { + "epoch": 0.05504587155963303, + "grad_norm": 7.2309651374816895, + "learning_rate": 2.690355329949238e-07, + "loss": 0.4537, + "mean_token_accuracy": 0.8348360061645508, + "num_tokens": 398774.0, + "step": 54 + }, + { + "epoch": 0.05606523955147808, + "grad_norm": 6.830613613128662, + "learning_rate": 2.7411167512690353e-07, + "loss": 0.4411, + "mean_token_accuracy": 0.8339318037033081, + "num_tokens": 407412.0, + "step": 55 + }, + { + "epoch": 0.05708460754332314, + "grad_norm": 7.2956461906433105, + "learning_rate": 2.7918781725888326e-07, + "loss": 0.4669, + "mean_token_accuracy": 0.8258403539657593, + "num_tokens": 414233.0, + "step": 56 + }, + { + "epoch": 0.0581039755351682, + "grad_norm": 7.630441665649414, + "learning_rate": 2.84263959390863e-07, + "loss": 0.4761, + "mean_token_accuracy": 0.8211209774017334, + "num_tokens": 421591.0, + "step": 57 + }, + { + "epoch": 0.059123343527013254, + "grad_norm": 7.995131015777588, + "learning_rate": 2.893401015228426e-07, + "loss": 0.4707, + "mean_token_accuracy": 0.8241152763366699, + "num_tokens": 428559.0, + "step": 58 + }, + { + "epoch": 0.060142711518858305, + "grad_norm": 9.10165023803711, + "learning_rate": 2.944162436548223e-07, + "loss": 0.4251, + "mean_token_accuracy": 0.8438158631324768, + "num_tokens": 435290.0, + "step": 59 + }, + { + "epoch": 0.06116207951070336, + "grad_norm": 8.083844184875488, + "learning_rate": 2.9949238578680204e-07, + "loss": 0.4682, + "mean_token_accuracy": 0.8238894939422607, + "num_tokens": 444312.0, + "step": 60 + }, + { + "epoch": 0.06218144750254842, + "grad_norm": 9.357006072998047, + "learning_rate": 3.0456852791878176e-07, + "loss": 0.437, + "mean_token_accuracy": 0.8342035412788391, + "num_tokens": 451502.0, + "step": 61 + }, + { + "epoch": 0.06320081549439348, + "grad_norm": 7.9630584716796875, + "learning_rate": 3.0964467005076137e-07, + "loss": 0.4052, + "mean_token_accuracy": 0.8484325408935547, + "num_tokens": 457957.0, + "step": 62 + }, + { + "epoch": 0.06422018348623854, + "grad_norm": 8.052801132202148, + "learning_rate": 3.147208121827411e-07, + "loss": 0.4286, + "mean_token_accuracy": 0.8355411291122437, + "num_tokens": 465042.0, + "step": 63 + }, + { + "epoch": 0.0652395514780836, + "grad_norm": 7.019420623779297, + "learning_rate": 3.197969543147208e-07, + "loss": 0.4063, + "mean_token_accuracy": 0.849837064743042, + "num_tokens": 473077.0, + "step": 64 + }, + { + "epoch": 0.06625891946992865, + "grad_norm": 8.610517501831055, + "learning_rate": 3.248730964467005e-07, + "loss": 0.4619, + "mean_token_accuracy": 0.8294556736946106, + "num_tokens": 479856.0, + "step": 65 + }, + { + "epoch": 0.0672782874617737, + "grad_norm": 6.911940574645996, + "learning_rate": 3.2994923857868015e-07, + "loss": 0.4145, + "mean_token_accuracy": 0.8434881567955017, + "num_tokens": 487287.0, + "step": 66 + }, + { + "epoch": 0.06829765545361875, + "grad_norm": 6.288229942321777, + "learning_rate": 3.350253807106599e-07, + "loss": 0.3894, + "mean_token_accuracy": 0.852220892906189, + "num_tokens": 495815.0, + "step": 67 + }, + { + "epoch": 0.06931702344546381, + "grad_norm": 6.280326843261719, + "learning_rate": 3.401015228426396e-07, + "loss": 0.4182, + "mean_token_accuracy": 0.8428940176963806, + "num_tokens": 503969.0, + "step": 68 + }, + { + "epoch": 0.07033639143730887, + "grad_norm": 7.168127536773682, + "learning_rate": 3.4517766497461926e-07, + "loss": 0.431, + "mean_token_accuracy": 0.844904363155365, + "num_tokens": 510858.0, + "step": 69 + }, + { + "epoch": 0.07135575942915393, + "grad_norm": 6.210775375366211, + "learning_rate": 3.50253807106599e-07, + "loss": 0.4317, + "mean_token_accuracy": 0.8423172235488892, + "num_tokens": 518343.0, + "step": 70 + }, + { + "epoch": 0.07237512742099898, + "grad_norm": 7.10784912109375, + "learning_rate": 3.5532994923857865e-07, + "loss": 0.4579, + "mean_token_accuracy": 0.8307130336761475, + "num_tokens": 524017.0, + "step": 71 + }, + { + "epoch": 0.07339449541284404, + "grad_norm": 5.968893051147461, + "learning_rate": 3.604060913705584e-07, + "loss": 0.3943, + "mean_token_accuracy": 0.8463324904441833, + "num_tokens": 531816.0, + "step": 72 + }, + { + "epoch": 0.0744138634046891, + "grad_norm": 5.974521160125732, + "learning_rate": 3.6548223350253804e-07, + "loss": 0.4444, + "mean_token_accuracy": 0.832889199256897, + "num_tokens": 539720.0, + "step": 73 + }, + { + "epoch": 0.07543323139653414, + "grad_norm": 6.43256950378418, + "learning_rate": 3.7055837563451777e-07, + "loss": 0.4113, + "mean_token_accuracy": 0.8442540168762207, + "num_tokens": 546122.0, + "step": 74 + }, + { + "epoch": 0.0764525993883792, + "grad_norm": 6.129632472991943, + "learning_rate": 3.7563451776649743e-07, + "loss": 0.3986, + "mean_token_accuracy": 0.8360404372215271, + "num_tokens": 552764.0, + "step": 75 + }, + { + "epoch": 0.07747196738022426, + "grad_norm": 5.329959869384766, + "learning_rate": 3.807106598984771e-07, + "loss": 0.4371, + "mean_token_accuracy": 0.832938015460968, + "num_tokens": 561731.0, + "step": 76 + }, + { + "epoch": 0.07849133537206932, + "grad_norm": 6.715664386749268, + "learning_rate": 3.857868020304568e-07, + "loss": 0.4154, + "mean_token_accuracy": 0.8349832892417908, + "num_tokens": 568309.0, + "step": 77 + }, + { + "epoch": 0.07951070336391437, + "grad_norm": 5.584300518035889, + "learning_rate": 3.9086294416243655e-07, + "loss": 0.3924, + "mean_token_accuracy": 0.8495333194732666, + "num_tokens": 575650.0, + "step": 78 + }, + { + "epoch": 0.08053007135575943, + "grad_norm": 6.195930004119873, + "learning_rate": 3.9593908629441627e-07, + "loss": 0.4064, + "mean_token_accuracy": 0.8412551879882812, + "num_tokens": 581199.0, + "step": 79 + }, + { + "epoch": 0.08154943934760449, + "grad_norm": 5.638856887817383, + "learning_rate": 4.010152284263959e-07, + "loss": 0.3923, + "mean_token_accuracy": 0.8543396592140198, + "num_tokens": 588910.0, + "step": 80 + }, + { + "epoch": 0.08256880733944955, + "grad_norm": 4.7872138023376465, + "learning_rate": 4.060913705583756e-07, + "loss": 0.3966, + "mean_token_accuracy": 0.8453965783119202, + "num_tokens": 597756.0, + "step": 81 + }, + { + "epoch": 0.0835881753312946, + "grad_norm": 6.197037696838379, + "learning_rate": 4.1116751269035533e-07, + "loss": 0.4343, + "mean_token_accuracy": 0.829355776309967, + "num_tokens": 604787.0, + "step": 82 + }, + { + "epoch": 0.08460754332313965, + "grad_norm": 5.587804794311523, + "learning_rate": 4.1624365482233505e-07, + "loss": 0.3731, + "mean_token_accuracy": 0.8562564849853516, + "num_tokens": 612535.0, + "step": 83 + }, + { + "epoch": 0.0856269113149847, + "grad_norm": 5.720602989196777, + "learning_rate": 4.2131979695431466e-07, + "loss": 0.4004, + "mean_token_accuracy": 0.8488187789916992, + "num_tokens": 619555.0, + "step": 84 + }, + { + "epoch": 0.08664627930682976, + "grad_norm": 5.473731994628906, + "learning_rate": 4.263959390862944e-07, + "loss": 0.3863, + "mean_token_accuracy": 0.8541125059127808, + "num_tokens": 627285.0, + "step": 85 + }, + { + "epoch": 0.08766564729867482, + "grad_norm": 7.066037654876709, + "learning_rate": 4.314720812182741e-07, + "loss": 0.4269, + "mean_token_accuracy": 0.8264515399932861, + "num_tokens": 632708.0, + "step": 86 + }, + { + "epoch": 0.08868501529051988, + "grad_norm": 4.943114757537842, + "learning_rate": 4.365482233502538e-07, + "loss": 0.3713, + "mean_token_accuracy": 0.8568798303604126, + "num_tokens": 640918.0, + "step": 87 + }, + { + "epoch": 0.08970438328236494, + "grad_norm": 5.34121036529541, + "learning_rate": 4.416243654822335e-07, + "loss": 0.3941, + "mean_token_accuracy": 0.8465211391448975, + "num_tokens": 649311.0, + "step": 88 + }, + { + "epoch": 0.09072375127421, + "grad_norm": 6.180456638336182, + "learning_rate": 4.4670050761421317e-07, + "loss": 0.4329, + "mean_token_accuracy": 0.8360943794250488, + "num_tokens": 655942.0, + "step": 89 + }, + { + "epoch": 0.09174311926605505, + "grad_norm": 6.2925896644592285, + "learning_rate": 4.517766497461929e-07, + "loss": 0.4478, + "mean_token_accuracy": 0.8297043442726135, + "num_tokens": 663172.0, + "step": 90 + }, + { + "epoch": 0.09276248725790011, + "grad_norm": 5.831687927246094, + "learning_rate": 4.5685279187817256e-07, + "loss": 0.4417, + "mean_token_accuracy": 0.8255857229232788, + "num_tokens": 671408.0, + "step": 91 + }, + { + "epoch": 0.09378185524974515, + "grad_norm": 6.2882890701293945, + "learning_rate": 4.619289340101523e-07, + "loss": 0.3721, + "mean_token_accuracy": 0.8565549254417419, + "num_tokens": 677760.0, + "step": 92 + }, + { + "epoch": 0.09480122324159021, + "grad_norm": 5.898828983306885, + "learning_rate": 4.6700507614213195e-07, + "loss": 0.461, + "mean_token_accuracy": 0.8180032968521118, + "num_tokens": 685209.0, + "step": 93 + }, + { + "epoch": 0.09582059123343527, + "grad_norm": 6.246304988861084, + "learning_rate": 4.7208121827411167e-07, + "loss": 0.4559, + "mean_token_accuracy": 0.8239033222198486, + "num_tokens": 691373.0, + "step": 94 + }, + { + "epoch": 0.09683995922528033, + "grad_norm": 5.911473274230957, + "learning_rate": 4.771573604060913e-07, + "loss": 0.4444, + "mean_token_accuracy": 0.8316642642021179, + "num_tokens": 698802.0, + "step": 95 + }, + { + "epoch": 0.09785932721712538, + "grad_norm": 6.868052005767822, + "learning_rate": 4.82233502538071e-07, + "loss": 0.3786, + "mean_token_accuracy": 0.8481454849243164, + "num_tokens": 705779.0, + "step": 96 + }, + { + "epoch": 0.09887869520897044, + "grad_norm": 5.679911136627197, + "learning_rate": 4.873096446700507e-07, + "loss": 0.3751, + "mean_token_accuracy": 0.8520702719688416, + "num_tokens": 712790.0, + "step": 97 + }, + { + "epoch": 0.0998980632008155, + "grad_norm": 5.296164512634277, + "learning_rate": 4.923857868020304e-07, + "loss": 0.4137, + "mean_token_accuracy": 0.8393490314483643, + "num_tokens": 719990.0, + "step": 98 + }, + { + "epoch": 0.10091743119266056, + "grad_norm": 6.001718997955322, + "learning_rate": 4.974619289340102e-07, + "loss": 0.4122, + "mean_token_accuracy": 0.840818464756012, + "num_tokens": 726328.0, + "step": 99 + }, + { + "epoch": 0.1019367991845056, + "grad_norm": 4.912004470825195, + "learning_rate": 5.025380710659898e-07, + "loss": 0.3829, + "mean_token_accuracy": 0.8490358591079712, + "num_tokens": 734812.0, + "step": 100 + }, + { + "epoch": 0.10295616717635066, + "grad_norm": 5.971344947814941, + "learning_rate": 5.076142131979695e-07, + "loss": 0.4589, + "mean_token_accuracy": 0.824362576007843, + "num_tokens": 742198.0, + "step": 101 + }, + { + "epoch": 0.10397553516819572, + "grad_norm": 7.538066387176514, + "learning_rate": 5.126903553299492e-07, + "loss": 0.4567, + "mean_token_accuracy": 0.8247807025909424, + "num_tokens": 748149.0, + "step": 102 + }, + { + "epoch": 0.10499490316004077, + "grad_norm": 5.558666706085205, + "learning_rate": 5.17766497461929e-07, + "loss": 0.4163, + "mean_token_accuracy": 0.8378702402114868, + "num_tokens": 755482.0, + "step": 103 + }, + { + "epoch": 0.10601427115188583, + "grad_norm": 7.142409801483154, + "learning_rate": 5.228426395939087e-07, + "loss": 0.4009, + "mean_token_accuracy": 0.8318968415260315, + "num_tokens": 761237.0, + "step": 104 + }, + { + "epoch": 0.10703363914373089, + "grad_norm": 5.8764328956604, + "learning_rate": 5.279187817258883e-07, + "loss": 0.4328, + "mean_token_accuracy": 0.8273512125015259, + "num_tokens": 768723.0, + "step": 105 + }, + { + "epoch": 0.10805300713557595, + "grad_norm": 5.614640235900879, + "learning_rate": 5.32994923857868e-07, + "loss": 0.4136, + "mean_token_accuracy": 0.8398871421813965, + "num_tokens": 775671.0, + "step": 106 + }, + { + "epoch": 0.109072375127421, + "grad_norm": 5.876361846923828, + "learning_rate": 5.380710659898476e-07, + "loss": 0.4133, + "mean_token_accuracy": 0.8334313631057739, + "num_tokens": 782573.0, + "step": 107 + }, + { + "epoch": 0.11009174311926606, + "grad_norm": 5.087253093719482, + "learning_rate": 5.431472081218273e-07, + "loss": 0.3969, + "mean_token_accuracy": 0.8452070951461792, + "num_tokens": 790554.0, + "step": 108 + }, + { + "epoch": 0.1111111111111111, + "grad_norm": 6.093214511871338, + "learning_rate": 5.482233502538071e-07, + "loss": 0.3964, + "mean_token_accuracy": 0.8426005244255066, + "num_tokens": 796844.0, + "step": 109 + }, + { + "epoch": 0.11213047910295616, + "grad_norm": 5.729161739349365, + "learning_rate": 5.532994923857868e-07, + "loss": 0.4097, + "mean_token_accuracy": 0.8442529439926147, + "num_tokens": 804053.0, + "step": 110 + }, + { + "epoch": 0.11314984709480122, + "grad_norm": 5.3184380531311035, + "learning_rate": 5.583756345177665e-07, + "loss": 0.3983, + "mean_token_accuracy": 0.8425759673118591, + "num_tokens": 812290.0, + "step": 111 + }, + { + "epoch": 0.11416921508664628, + "grad_norm": 5.865428924560547, + "learning_rate": 5.634517766497462e-07, + "loss": 0.4382, + "mean_token_accuracy": 0.8334693312644958, + "num_tokens": 819351.0, + "step": 112 + }, + { + "epoch": 0.11518858307849134, + "grad_norm": 5.347846031188965, + "learning_rate": 5.68527918781726e-07, + "loss": 0.3619, + "mean_token_accuracy": 0.851229190826416, + "num_tokens": 826750.0, + "step": 113 + }, + { + "epoch": 0.1162079510703364, + "grad_norm": 5.3884596824646, + "learning_rate": 5.736040609137056e-07, + "loss": 0.4144, + "mean_token_accuracy": 0.8392070531845093, + "num_tokens": 835212.0, + "step": 114 + }, + { + "epoch": 0.11722731906218145, + "grad_norm": 7.289360046386719, + "learning_rate": 5.786802030456852e-07, + "loss": 0.4096, + "mean_token_accuracy": 0.8401811122894287, + "num_tokens": 841360.0, + "step": 115 + }, + { + "epoch": 0.11824668705402651, + "grad_norm": 5.759810447692871, + "learning_rate": 5.837563451776649e-07, + "loss": 0.4313, + "mean_token_accuracy": 0.829369068145752, + "num_tokens": 849008.0, + "step": 116 + }, + { + "epoch": 0.11926605504587157, + "grad_norm": 6.401013374328613, + "learning_rate": 5.888324873096446e-07, + "loss": 0.4386, + "mean_token_accuracy": 0.822128415107727, + "num_tokens": 855980.0, + "step": 117 + }, + { + "epoch": 0.12028542303771661, + "grad_norm": 6.396369934082031, + "learning_rate": 5.939086294416243e-07, + "loss": 0.385, + "mean_token_accuracy": 0.8483094573020935, + "num_tokens": 863324.0, + "step": 118 + }, + { + "epoch": 0.12130479102956167, + "grad_norm": 7.0065202713012695, + "learning_rate": 5.989847715736041e-07, + "loss": 0.4507, + "mean_token_accuracy": 0.8292804956436157, + "num_tokens": 869087.0, + "step": 119 + }, + { + "epoch": 0.12232415902140673, + "grad_norm": 5.3116374015808105, + "learning_rate": 6.040609137055838e-07, + "loss": 0.3845, + "mean_token_accuracy": 0.8503082990646362, + "num_tokens": 876490.0, + "step": 120 + }, + { + "epoch": 0.12334352701325178, + "grad_norm": 6.009892463684082, + "learning_rate": 6.091370558375635e-07, + "loss": 0.4053, + "mean_token_accuracy": 0.8306334614753723, + "num_tokens": 883183.0, + "step": 121 + }, + { + "epoch": 0.12436289500509684, + "grad_norm": 5.236307144165039, + "learning_rate": 6.14213197969543e-07, + "loss": 0.4226, + "mean_token_accuracy": 0.8338778614997864, + "num_tokens": 890758.0, + "step": 122 + }, + { + "epoch": 0.12538226299694188, + "grad_norm": 6.348887920379639, + "learning_rate": 6.192893401015227e-07, + "loss": 0.4708, + "mean_token_accuracy": 0.8263953924179077, + "num_tokens": 897738.0, + "step": 123 + }, + { + "epoch": 0.12640163098878696, + "grad_norm": 4.96765661239624, + "learning_rate": 6.243654822335025e-07, + "loss": 0.3841, + "mean_token_accuracy": 0.8475692868232727, + "num_tokens": 905432.0, + "step": 124 + }, + { + "epoch": 0.127420998980632, + "grad_norm": 6.044402599334717, + "learning_rate": 6.294416243654822e-07, + "loss": 0.421, + "mean_token_accuracy": 0.8305132389068604, + "num_tokens": 911314.0, + "step": 125 + }, + { + "epoch": 0.12844036697247707, + "grad_norm": 5.082648754119873, + "learning_rate": 6.345177664974619e-07, + "loss": 0.4043, + "mean_token_accuracy": 0.8429370522499084, + "num_tokens": 919548.0, + "step": 126 + }, + { + "epoch": 0.12945973496432212, + "grad_norm": 6.187060832977295, + "learning_rate": 6.395939086294416e-07, + "loss": 0.4162, + "mean_token_accuracy": 0.8429096937179565, + "num_tokens": 926331.0, + "step": 127 + }, + { + "epoch": 0.1304791029561672, + "grad_norm": 6.308645725250244, + "learning_rate": 6.446700507614214e-07, + "loss": 0.4049, + "mean_token_accuracy": 0.8381337523460388, + "num_tokens": 932142.0, + "step": 128 + }, + { + "epoch": 0.13149847094801223, + "grad_norm": 6.367486000061035, + "learning_rate": 6.49746192893401e-07, + "loss": 0.4245, + "mean_token_accuracy": 0.8324413299560547, + "num_tokens": 938945.0, + "step": 129 + }, + { + "epoch": 0.1325178389398573, + "grad_norm": 5.1081109046936035, + "learning_rate": 6.548223350253807e-07, + "loss": 0.3566, + "mean_token_accuracy": 0.8596411943435669, + "num_tokens": 946344.0, + "step": 130 + }, + { + "epoch": 0.13353720693170235, + "grad_norm": 6.158345699310303, + "learning_rate": 6.598984771573603e-07, + "loss": 0.3944, + "mean_token_accuracy": 0.8488456606864929, + "num_tokens": 953019.0, + "step": 131 + }, + { + "epoch": 0.1345565749235474, + "grad_norm": 5.861425876617432, + "learning_rate": 6.6497461928934e-07, + "loss": 0.3846, + "mean_token_accuracy": 0.8522497415542603, + "num_tokens": 959912.0, + "step": 132 + }, + { + "epoch": 0.13557594291539246, + "grad_norm": 5.815319538116455, + "learning_rate": 6.700507614213197e-07, + "loss": 0.4217, + "mean_token_accuracy": 0.8354474306106567, + "num_tokens": 967887.0, + "step": 133 + }, + { + "epoch": 0.1365953109072375, + "grad_norm": 7.1927642822265625, + "learning_rate": 6.751269035532995e-07, + "loss": 0.4224, + "mean_token_accuracy": 0.8303354382514954, + "num_tokens": 973625.0, + "step": 134 + }, + { + "epoch": 0.13761467889908258, + "grad_norm": 5.8909220695495605, + "learning_rate": 6.802030456852792e-07, + "loss": 0.4269, + "mean_token_accuracy": 0.8325914144515991, + "num_tokens": 979980.0, + "step": 135 + }, + { + "epoch": 0.13863404689092762, + "grad_norm": 6.807290077209473, + "learning_rate": 6.852791878172588e-07, + "loss": 0.4716, + "mean_token_accuracy": 0.8122974634170532, + "num_tokens": 986571.0, + "step": 136 + }, + { + "epoch": 0.1396534148827727, + "grad_norm": 6.602284908294678, + "learning_rate": 6.903553299492385e-07, + "loss": 0.4318, + "mean_token_accuracy": 0.8298578262329102, + "num_tokens": 993245.0, + "step": 137 + }, + { + "epoch": 0.14067278287461774, + "grad_norm": 6.881693363189697, + "learning_rate": 6.954314720812183e-07, + "loss": 0.4368, + "mean_token_accuracy": 0.8297092318534851, + "num_tokens": 999009.0, + "step": 138 + }, + { + "epoch": 0.14169215086646278, + "grad_norm": 6.200334072113037, + "learning_rate": 7.00507614213198e-07, + "loss": 0.4167, + "mean_token_accuracy": 0.8453640937805176, + "num_tokens": 1006172.0, + "step": 139 + }, + { + "epoch": 0.14271151885830785, + "grad_norm": 6.743097305297852, + "learning_rate": 7.055837563451776e-07, + "loss": 0.4095, + "mean_token_accuracy": 0.8401176929473877, + "num_tokens": 1012752.0, + "step": 140 + }, + { + "epoch": 0.1437308868501529, + "grad_norm": 4.933243274688721, + "learning_rate": 7.106598984771573e-07, + "loss": 0.3944, + "mean_token_accuracy": 0.838455319404602, + "num_tokens": 1020606.0, + "step": 141 + }, + { + "epoch": 0.14475025484199797, + "grad_norm": 4.7325334548950195, + "learning_rate": 7.15736040609137e-07, + "loss": 0.3809, + "mean_token_accuracy": 0.8526402711868286, + "num_tokens": 1030213.0, + "step": 142 + }, + { + "epoch": 0.145769622833843, + "grad_norm": 5.239090442657471, + "learning_rate": 7.208121827411168e-07, + "loss": 0.387, + "mean_token_accuracy": 0.8424021601676941, + "num_tokens": 1037698.0, + "step": 143 + }, + { + "epoch": 0.14678899082568808, + "grad_norm": 5.733617782592773, + "learning_rate": 7.258883248730964e-07, + "loss": 0.4107, + "mean_token_accuracy": 0.8400851488113403, + "num_tokens": 1045200.0, + "step": 144 + }, + { + "epoch": 0.14780835881753313, + "grad_norm": 5.765109062194824, + "learning_rate": 7.309644670050761e-07, + "loss": 0.4147, + "mean_token_accuracy": 0.8384674787521362, + "num_tokens": 1052411.0, + "step": 145 + }, + { + "epoch": 0.1488277268093782, + "grad_norm": 5.483977794647217, + "learning_rate": 7.360406091370558e-07, + "loss": 0.3882, + "mean_token_accuracy": 0.8396845459938049, + "num_tokens": 1059595.0, + "step": 146 + }, + { + "epoch": 0.14984709480122324, + "grad_norm": 5.562849998474121, + "learning_rate": 7.411167512690355e-07, + "loss": 0.4168, + "mean_token_accuracy": 0.827721118927002, + "num_tokens": 1066930.0, + "step": 147 + }, + { + "epoch": 0.15086646279306828, + "grad_norm": 6.511322498321533, + "learning_rate": 7.461928934010153e-07, + "loss": 0.4143, + "mean_token_accuracy": 0.8414583802223206, + "num_tokens": 1073269.0, + "step": 148 + }, + { + "epoch": 0.15188583078491336, + "grad_norm": 6.648353099822998, + "learning_rate": 7.512690355329949e-07, + "loss": 0.3813, + "mean_token_accuracy": 0.8475778102874756, + "num_tokens": 1079271.0, + "step": 149 + }, + { + "epoch": 0.1529051987767584, + "grad_norm": 5.806103229522705, + "learning_rate": 7.563451776649746e-07, + "loss": 0.4585, + "mean_token_accuracy": 0.8185784816741943, + "num_tokens": 1086221.0, + "step": 150 + }, + { + "epoch": 0.15392456676860347, + "grad_norm": 6.262481212615967, + "learning_rate": 7.614213197969542e-07, + "loss": 0.3945, + "mean_token_accuracy": 0.85139000415802, + "num_tokens": 1093021.0, + "step": 151 + }, + { + "epoch": 0.15494393476044852, + "grad_norm": 4.893447399139404, + "learning_rate": 7.664974619289339e-07, + "loss": 0.3653, + "mean_token_accuracy": 0.8596205711364746, + "num_tokens": 1100480.0, + "step": 152 + }, + { + "epoch": 0.1559633027522936, + "grad_norm": 5.208713531494141, + "learning_rate": 7.715736040609137e-07, + "loss": 0.403, + "mean_token_accuracy": 0.846179723739624, + "num_tokens": 1107962.0, + "step": 153 + }, + { + "epoch": 0.15698267074413863, + "grad_norm": 5.685441493988037, + "learning_rate": 7.766497461928934e-07, + "loss": 0.3525, + "mean_token_accuracy": 0.8554806113243103, + "num_tokens": 1114713.0, + "step": 154 + }, + { + "epoch": 0.1580020387359837, + "grad_norm": 5.248551368713379, + "learning_rate": 7.817258883248731e-07, + "loss": 0.3717, + "mean_token_accuracy": 0.8574260473251343, + "num_tokens": 1122404.0, + "step": 155 + }, + { + "epoch": 0.15902140672782875, + "grad_norm": 5.200737953186035, + "learning_rate": 7.868020304568528e-07, + "loss": 0.4001, + "mean_token_accuracy": 0.8435255289077759, + "num_tokens": 1130125.0, + "step": 156 + }, + { + "epoch": 0.1600407747196738, + "grad_norm": 6.437376499176025, + "learning_rate": 7.918781725888325e-07, + "loss": 0.4286, + "mean_token_accuracy": 0.8351315259933472, + "num_tokens": 1136557.0, + "step": 157 + }, + { + "epoch": 0.16106014271151886, + "grad_norm": 5.641631603240967, + "learning_rate": 7.96954314720812e-07, + "loss": 0.3834, + "mean_token_accuracy": 0.8541744351387024, + "num_tokens": 1144392.0, + "step": 158 + }, + { + "epoch": 0.1620795107033639, + "grad_norm": 5.477780342102051, + "learning_rate": 8.020304568527918e-07, + "loss": 0.4082, + "mean_token_accuracy": 0.843484103679657, + "num_tokens": 1151230.0, + "step": 159 + }, + { + "epoch": 0.16309887869520898, + "grad_norm": 5.401800155639648, + "learning_rate": 8.071065989847715e-07, + "loss": 0.3438, + "mean_token_accuracy": 0.8722546100616455, + "num_tokens": 1157775.0, + "step": 160 + }, + { + "epoch": 0.16411824668705402, + "grad_norm": 5.384525775909424, + "learning_rate": 8.121827411167512e-07, + "loss": 0.3949, + "mean_token_accuracy": 0.8473877906799316, + "num_tokens": 1165965.0, + "step": 161 + }, + { + "epoch": 0.1651376146788991, + "grad_norm": 5.223589897155762, + "learning_rate": 8.172588832487309e-07, + "loss": 0.4093, + "mean_token_accuracy": 0.8391587138175964, + "num_tokens": 1173730.0, + "step": 162 + }, + { + "epoch": 0.16615698267074414, + "grad_norm": 5.427058696746826, + "learning_rate": 8.223350253807107e-07, + "loss": 0.3841, + "mean_token_accuracy": 0.8476698398590088, + "num_tokens": 1181498.0, + "step": 163 + }, + { + "epoch": 0.1671763506625892, + "grad_norm": 6.167184352874756, + "learning_rate": 8.274111675126904e-07, + "loss": 0.3873, + "mean_token_accuracy": 0.8481369018554688, + "num_tokens": 1187577.0, + "step": 164 + }, + { + "epoch": 0.16819571865443425, + "grad_norm": 6.003235816955566, + "learning_rate": 8.324873096446701e-07, + "loss": 0.4021, + "mean_token_accuracy": 0.8406115770339966, + "num_tokens": 1194678.0, + "step": 165 + }, + { + "epoch": 0.1692150866462793, + "grad_norm": 5.543849468231201, + "learning_rate": 8.375634517766497e-07, + "loss": 0.3793, + "mean_token_accuracy": 0.852969765663147, + "num_tokens": 1201372.0, + "step": 166 + }, + { + "epoch": 0.17023445463812437, + "grad_norm": 5.975922584533691, + "learning_rate": 8.426395939086293e-07, + "loss": 0.4012, + "mean_token_accuracy": 0.8438053727149963, + "num_tokens": 1207731.0, + "step": 167 + }, + { + "epoch": 0.1712538226299694, + "grad_norm": 7.375309467315674, + "learning_rate": 8.47715736040609e-07, + "loss": 0.4392, + "mean_token_accuracy": 0.8293983936309814, + "num_tokens": 1213190.0, + "step": 168 + }, + { + "epoch": 0.17227319062181448, + "grad_norm": 4.566862106323242, + "learning_rate": 8.527918781725888e-07, + "loss": 0.3593, + "mean_token_accuracy": 0.8596283197402954, + "num_tokens": 1222099.0, + "step": 169 + }, + { + "epoch": 0.17329255861365953, + "grad_norm": 5.791296005249023, + "learning_rate": 8.578680203045685e-07, + "loss": 0.4031, + "mean_token_accuracy": 0.8439279198646545, + "num_tokens": 1229076.0, + "step": 170 + }, + { + "epoch": 0.1743119266055046, + "grad_norm": 6.166365623474121, + "learning_rate": 8.629441624365482e-07, + "loss": 0.3602, + "mean_token_accuracy": 0.8569746613502502, + "num_tokens": 1235682.0, + "step": 171 + }, + { + "epoch": 0.17533129459734964, + "grad_norm": 5.5373687744140625, + "learning_rate": 8.680203045685279e-07, + "loss": 0.4094, + "mean_token_accuracy": 0.838245153427124, + "num_tokens": 1243911.0, + "step": 172 + }, + { + "epoch": 0.1763506625891947, + "grad_norm": 7.255425930023193, + "learning_rate": 8.730964467005076e-07, + "loss": 0.4469, + "mean_token_accuracy": 0.8210601806640625, + "num_tokens": 1249212.0, + "step": 173 + }, + { + "epoch": 0.17737003058103976, + "grad_norm": 5.328028202056885, + "learning_rate": 8.781725888324873e-07, + "loss": 0.3905, + "mean_token_accuracy": 0.8383325338363647, + "num_tokens": 1256985.0, + "step": 174 + }, + { + "epoch": 0.1783893985728848, + "grad_norm": 5.774742126464844, + "learning_rate": 8.83248730964467e-07, + "loss": 0.4119, + "mean_token_accuracy": 0.8427859544754028, + "num_tokens": 1263141.0, + "step": 175 + }, + { + "epoch": 0.17940876656472987, + "grad_norm": 6.340662479400635, + "learning_rate": 8.883248730964466e-07, + "loss": 0.436, + "mean_token_accuracy": 0.830388069152832, + "num_tokens": 1269525.0, + "step": 176 + }, + { + "epoch": 0.18042813455657492, + "grad_norm": 5.8206915855407715, + "learning_rate": 8.934010152284263e-07, + "loss": 0.4093, + "mean_token_accuracy": 0.845696210861206, + "num_tokens": 1276149.0, + "step": 177 + }, + { + "epoch": 0.18144750254842, + "grad_norm": 7.688342094421387, + "learning_rate": 8.984771573604061e-07, + "loss": 0.4358, + "mean_token_accuracy": 0.8318321704864502, + "num_tokens": 1281793.0, + "step": 178 + }, + { + "epoch": 0.18246687054026503, + "grad_norm": 6.036313056945801, + "learning_rate": 9.035532994923858e-07, + "loss": 0.3982, + "mean_token_accuracy": 0.8422325849533081, + "num_tokens": 1288018.0, + "step": 179 + }, + { + "epoch": 0.1834862385321101, + "grad_norm": 5.3137383460998535, + "learning_rate": 9.086294416243654e-07, + "loss": 0.3535, + "mean_token_accuracy": 0.8653435111045837, + "num_tokens": 1295326.0, + "step": 180 + }, + { + "epoch": 0.18450560652395515, + "grad_norm": 5.368961811065674, + "learning_rate": 9.137055837563451e-07, + "loss": 0.3868, + "mean_token_accuracy": 0.8429610729217529, + "num_tokens": 1302470.0, + "step": 181 + }, + { + "epoch": 0.18552497451580022, + "grad_norm": 5.838701248168945, + "learning_rate": 9.187817258883248e-07, + "loss": 0.4442, + "mean_token_accuracy": 0.8295447826385498, + "num_tokens": 1310084.0, + "step": 182 + }, + { + "epoch": 0.18654434250764526, + "grad_norm": 5.4157514572143555, + "learning_rate": 9.238578680203046e-07, + "loss": 0.3888, + "mean_token_accuracy": 0.8430968523025513, + "num_tokens": 1317590.0, + "step": 183 + }, + { + "epoch": 0.1875637104994903, + "grad_norm": 6.23513650894165, + "learning_rate": 9.289340101522843e-07, + "loss": 0.3798, + "mean_token_accuracy": 0.8454179763793945, + "num_tokens": 1324272.0, + "step": 184 + }, + { + "epoch": 0.18858307849133538, + "grad_norm": 5.947885990142822, + "learning_rate": 9.340101522842639e-07, + "loss": 0.3936, + "mean_token_accuracy": 0.8410367965698242, + "num_tokens": 1330577.0, + "step": 185 + }, + { + "epoch": 0.18960244648318042, + "grad_norm": 6.264049530029297, + "learning_rate": 9.390862944162436e-07, + "loss": 0.4026, + "mean_token_accuracy": 0.8448988199234009, + "num_tokens": 1336966.0, + "step": 186 + }, + { + "epoch": 0.1906218144750255, + "grad_norm": 5.035786151885986, + "learning_rate": 9.441624365482233e-07, + "loss": 0.4094, + "mean_token_accuracy": 0.8403424620628357, + "num_tokens": 1345442.0, + "step": 187 + }, + { + "epoch": 0.19164118246687054, + "grad_norm": 5.972903728485107, + "learning_rate": 9.49238578680203e-07, + "loss": 0.4019, + "mean_token_accuracy": 0.8459986448287964, + "num_tokens": 1352777.0, + "step": 188 + }, + { + "epoch": 0.1926605504587156, + "grad_norm": 5.3063435554504395, + "learning_rate": 9.543147208121826e-07, + "loss": 0.3926, + "mean_token_accuracy": 0.8509308099746704, + "num_tokens": 1360038.0, + "step": 189 + }, + { + "epoch": 0.19367991845056065, + "grad_norm": 6.072782039642334, + "learning_rate": 9.593908629441624e-07, + "loss": 0.4037, + "mean_token_accuracy": 0.8413677215576172, + "num_tokens": 1366421.0, + "step": 190 + }, + { + "epoch": 0.1946992864424057, + "grad_norm": 6.250436305999756, + "learning_rate": 9.64467005076142e-07, + "loss": 0.3898, + "mean_token_accuracy": 0.845289945602417, + "num_tokens": 1372615.0, + "step": 191 + }, + { + "epoch": 0.19571865443425077, + "grad_norm": 5.0199666023254395, + "learning_rate": 9.695431472081218e-07, + "loss": 0.3604, + "mean_token_accuracy": 0.8609315752983093, + "num_tokens": 1380754.0, + "step": 192 + }, + { + "epoch": 0.1967380224260958, + "grad_norm": 5.768274307250977, + "learning_rate": 9.746192893401015e-07, + "loss": 0.4321, + "mean_token_accuracy": 0.8283120393753052, + "num_tokens": 1388074.0, + "step": 193 + }, + { + "epoch": 0.19775739041794088, + "grad_norm": 5.066155433654785, + "learning_rate": 9.796954314720813e-07, + "loss": 0.3819, + "mean_token_accuracy": 0.8517265915870667, + "num_tokens": 1396703.0, + "step": 194 + }, + { + "epoch": 0.19877675840978593, + "grad_norm": 5.272959232330322, + "learning_rate": 9.84771573604061e-07, + "loss": 0.3297, + "mean_token_accuracy": 0.8563237190246582, + "num_tokens": 1403698.0, + "step": 195 + }, + { + "epoch": 0.199796126401631, + "grad_norm": 5.499087333679199, + "learning_rate": 9.898477157360405e-07, + "loss": 0.393, + "mean_token_accuracy": 0.8408558368682861, + "num_tokens": 1412333.0, + "step": 196 + }, + { + "epoch": 0.20081549439347604, + "grad_norm": 5.107669353485107, + "learning_rate": 9.949238578680203e-07, + "loss": 0.3758, + "mean_token_accuracy": 0.8485940098762512, + "num_tokens": 1420083.0, + "step": 197 + }, + { + "epoch": 0.2018348623853211, + "grad_norm": 5.928592205047607, + "learning_rate": 1e-06, + "loss": 0.4285, + "mean_token_accuracy": 0.8390728831291199, + "num_tokens": 1426368.0, + "step": 198 + }, + { + "epoch": 0.20285423037716616, + "grad_norm": 4.525904178619385, + "learning_rate": 1e-06, + "loss": 0.3702, + "mean_token_accuracy": 0.8519665002822876, + "num_tokens": 1435296.0, + "step": 199 + }, + { + "epoch": 0.2038735983690112, + "grad_norm": 4.157371997833252, + "learning_rate": 1e-06, + "loss": 0.3587, + "mean_token_accuracy": 0.8562180995941162, + "num_tokens": 1444893.0, + "step": 200 + }, + { + "epoch": 0.20489296636085627, + "grad_norm": 5.335367679595947, + "learning_rate": 1e-06, + "loss": 0.3843, + "mean_token_accuracy": 0.8466694951057434, + "num_tokens": 1452231.0, + "step": 201 + }, + { + "epoch": 0.20591233435270132, + "grad_norm": 6.835778713226318, + "learning_rate": 1e-06, + "loss": 0.429, + "mean_token_accuracy": 0.8311684727668762, + "num_tokens": 1458342.0, + "step": 202 + }, + { + "epoch": 0.2069317023445464, + "grad_norm": 5.293552398681641, + "learning_rate": 1e-06, + "loss": 0.4013, + "mean_token_accuracy": 0.8372024297714233, + "num_tokens": 1465818.0, + "step": 203 + }, + { + "epoch": 0.20795107033639143, + "grad_norm": 6.596436500549316, + "learning_rate": 1e-06, + "loss": 0.3913, + "mean_token_accuracy": 0.8483570218086243, + "num_tokens": 1471937.0, + "step": 204 + }, + { + "epoch": 0.2089704383282365, + "grad_norm": 5.823583602905273, + "learning_rate": 1e-06, + "loss": 0.4019, + "mean_token_accuracy": 0.8314054012298584, + "num_tokens": 1478937.0, + "step": 205 + }, + { + "epoch": 0.20998980632008155, + "grad_norm": 5.4598469734191895, + "learning_rate": 1e-06, + "loss": 0.4019, + "mean_token_accuracy": 0.8455653190612793, + "num_tokens": 1486194.0, + "step": 206 + }, + { + "epoch": 0.21100917431192662, + "grad_norm": 6.418130874633789, + "learning_rate": 1e-06, + "loss": 0.3938, + "mean_token_accuracy": 0.8404214382171631, + "num_tokens": 1493441.0, + "step": 207 + }, + { + "epoch": 0.21202854230377166, + "grad_norm": 5.07489538192749, + "learning_rate": 1e-06, + "loss": 0.3309, + "mean_token_accuracy": 0.855706512928009, + "num_tokens": 1501036.0, + "step": 208 + }, + { + "epoch": 0.2130479102956167, + "grad_norm": 6.035568714141846, + "learning_rate": 1e-06, + "loss": 0.3869, + "mean_token_accuracy": 0.8432818055152893, + "num_tokens": 1507416.0, + "step": 209 + }, + { + "epoch": 0.21406727828746178, + "grad_norm": 6.9841790199279785, + "learning_rate": 1e-06, + "loss": 0.4134, + "mean_token_accuracy": 0.8403686881065369, + "num_tokens": 1513757.0, + "step": 210 + }, + { + "epoch": 0.21508664627930682, + "grad_norm": 5.362586498260498, + "learning_rate": 1e-06, + "loss": 0.4304, + "mean_token_accuracy": 0.8298053741455078, + "num_tokens": 1521581.0, + "step": 211 + }, + { + "epoch": 0.2161060142711519, + "grad_norm": 7.383844375610352, + "learning_rate": 1e-06, + "loss": 0.3912, + "mean_token_accuracy": 0.8498966097831726, + "num_tokens": 1527659.0, + "step": 212 + }, + { + "epoch": 0.21712538226299694, + "grad_norm": 5.408533096313477, + "learning_rate": 1e-06, + "loss": 0.3984, + "mean_token_accuracy": 0.8436784744262695, + "num_tokens": 1534973.0, + "step": 213 + }, + { + "epoch": 0.218144750254842, + "grad_norm": 5.730254173278809, + "learning_rate": 1e-06, + "loss": 0.3741, + "mean_token_accuracy": 0.849095344543457, + "num_tokens": 1541701.0, + "step": 214 + }, + { + "epoch": 0.21916411824668705, + "grad_norm": 6.127647399902344, + "learning_rate": 1e-06, + "loss": 0.4239, + "mean_token_accuracy": 0.833275318145752, + "num_tokens": 1548215.0, + "step": 215 + }, + { + "epoch": 0.22018348623853212, + "grad_norm": 5.306388854980469, + "learning_rate": 1e-06, + "loss": 0.4027, + "mean_token_accuracy": 0.8393712639808655, + "num_tokens": 1556367.0, + "step": 216 + }, + { + "epoch": 0.22120285423037717, + "grad_norm": 7.4842610359191895, + "learning_rate": 1e-06, + "loss": 0.4261, + "mean_token_accuracy": 0.833489716053009, + "num_tokens": 1562062.0, + "step": 217 + }, + { + "epoch": 0.2222222222222222, + "grad_norm": 5.099609375, + "learning_rate": 1e-06, + "loss": 0.3698, + "mean_token_accuracy": 0.8527371883392334, + "num_tokens": 1570013.0, + "step": 218 + }, + { + "epoch": 0.22324159021406728, + "grad_norm": 4.966230869293213, + "learning_rate": 1e-06, + "loss": 0.3944, + "mean_token_accuracy": 0.8444109559059143, + "num_tokens": 1578273.0, + "step": 219 + }, + { + "epoch": 0.22426095820591233, + "grad_norm": 5.105199337005615, + "learning_rate": 1e-06, + "loss": 0.3798, + "mean_token_accuracy": 0.8510867953300476, + "num_tokens": 1586237.0, + "step": 220 + }, + { + "epoch": 0.2252803261977574, + "grad_norm": 6.531620025634766, + "learning_rate": 1e-06, + "loss": 0.4446, + "mean_token_accuracy": 0.8315724730491638, + "num_tokens": 1592503.0, + "step": 221 + }, + { + "epoch": 0.22629969418960244, + "grad_norm": 6.357755184173584, + "learning_rate": 1e-06, + "loss": 0.3865, + "mean_token_accuracy": 0.8480254411697388, + "num_tokens": 1600339.0, + "step": 222 + }, + { + "epoch": 0.2273190621814475, + "grad_norm": 5.8069329261779785, + "learning_rate": 1e-06, + "loss": 0.4233, + "mean_token_accuracy": 0.8400685787200928, + "num_tokens": 1608089.0, + "step": 223 + }, + { + "epoch": 0.22833843017329256, + "grad_norm": 6.484148979187012, + "learning_rate": 1e-06, + "loss": 0.3915, + "mean_token_accuracy": 0.8452358245849609, + "num_tokens": 1614168.0, + "step": 224 + }, + { + "epoch": 0.22935779816513763, + "grad_norm": 6.658324241638184, + "learning_rate": 1e-06, + "loss": 0.4331, + "mean_token_accuracy": 0.8224902749061584, + "num_tokens": 1620123.0, + "step": 225 + }, + { + "epoch": 0.23037716615698267, + "grad_norm": 6.394359111785889, + "learning_rate": 1e-06, + "loss": 0.4298, + "mean_token_accuracy": 0.8314876556396484, + "num_tokens": 1626724.0, + "step": 226 + }, + { + "epoch": 0.23139653414882771, + "grad_norm": 5.003575801849365, + "learning_rate": 1e-06, + "loss": 0.3634, + "mean_token_accuracy": 0.8605597019195557, + "num_tokens": 1635319.0, + "step": 227 + }, + { + "epoch": 0.2324159021406728, + "grad_norm": 5.419402599334717, + "learning_rate": 1e-06, + "loss": 0.398, + "mean_token_accuracy": 0.8337389826774597, + "num_tokens": 1642499.0, + "step": 228 + }, + { + "epoch": 0.23343527013251783, + "grad_norm": 6.842245101928711, + "learning_rate": 1e-06, + "loss": 0.3938, + "mean_token_accuracy": 0.8465895056724548, + "num_tokens": 1648453.0, + "step": 229 + }, + { + "epoch": 0.2344546381243629, + "grad_norm": 4.691684722900391, + "learning_rate": 1e-06, + "loss": 0.3981, + "mean_token_accuracy": 0.8425484299659729, + "num_tokens": 1657399.0, + "step": 230 + }, + { + "epoch": 0.23547400611620795, + "grad_norm": 5.062182426452637, + "learning_rate": 1e-06, + "loss": 0.3849, + "mean_token_accuracy": 0.8526610136032104, + "num_tokens": 1665741.0, + "step": 231 + }, + { + "epoch": 0.23649337410805302, + "grad_norm": 5.80242919921875, + "learning_rate": 1e-06, + "loss": 0.4148, + "mean_token_accuracy": 0.8386773467063904, + "num_tokens": 1673106.0, + "step": 232 + }, + { + "epoch": 0.23751274209989806, + "grad_norm": 5.381566524505615, + "learning_rate": 1e-06, + "loss": 0.4018, + "mean_token_accuracy": 0.842765212059021, + "num_tokens": 1680549.0, + "step": 233 + }, + { + "epoch": 0.23853211009174313, + "grad_norm": 6.509912490844727, + "learning_rate": 1e-06, + "loss": 0.4215, + "mean_token_accuracy": 0.8327610492706299, + "num_tokens": 1686870.0, + "step": 234 + }, + { + "epoch": 0.23955147808358818, + "grad_norm": 5.044449806213379, + "learning_rate": 1e-06, + "loss": 0.3896, + "mean_token_accuracy": 0.852663516998291, + "num_tokens": 1695473.0, + "step": 235 + }, + { + "epoch": 0.24057084607543322, + "grad_norm": 5.406460285186768, + "learning_rate": 1e-06, + "loss": 0.4021, + "mean_token_accuracy": 0.8445947170257568, + "num_tokens": 1702840.0, + "step": 236 + }, + { + "epoch": 0.2415902140672783, + "grad_norm": 5.808995246887207, + "learning_rate": 1e-06, + "loss": 0.4224, + "mean_token_accuracy": 0.8424966335296631, + "num_tokens": 1709540.0, + "step": 237 + }, + { + "epoch": 0.24260958205912334, + "grad_norm": 5.5747456550598145, + "learning_rate": 1e-06, + "loss": 0.4032, + "mean_token_accuracy": 0.8451700210571289, + "num_tokens": 1716333.0, + "step": 238 + }, + { + "epoch": 0.2436289500509684, + "grad_norm": 6.171189785003662, + "learning_rate": 1e-06, + "loss": 0.4307, + "mean_token_accuracy": 0.8413604497909546, + "num_tokens": 1723958.0, + "step": 239 + }, + { + "epoch": 0.24464831804281345, + "grad_norm": 5.566725730895996, + "learning_rate": 1e-06, + "loss": 0.4221, + "mean_token_accuracy": 0.8284014463424683, + "num_tokens": 1730601.0, + "step": 240 + }, + { + "epoch": 0.24566768603465852, + "grad_norm": 5.358464241027832, + "learning_rate": 1e-06, + "loss": 0.3792, + "mean_token_accuracy": 0.8378663063049316, + "num_tokens": 1737737.0, + "step": 241 + }, + { + "epoch": 0.24668705402650357, + "grad_norm": 6.003946304321289, + "learning_rate": 1e-06, + "loss": 0.3709, + "mean_token_accuracy": 0.8526804447174072, + "num_tokens": 1744085.0, + "step": 242 + }, + { + "epoch": 0.24770642201834864, + "grad_norm": 6.205704689025879, + "learning_rate": 1e-06, + "loss": 0.3709, + "mean_token_accuracy": 0.8541709184646606, + "num_tokens": 1750125.0, + "step": 243 + }, + { + "epoch": 0.24872579001019368, + "grad_norm": 5.6227216720581055, + "learning_rate": 1e-06, + "loss": 0.3785, + "mean_token_accuracy": 0.8509168028831482, + "num_tokens": 1758063.0, + "step": 244 + }, + { + "epoch": 0.24974515800203873, + "grad_norm": 5.729758262634277, + "learning_rate": 1e-06, + "loss": 0.4085, + "mean_token_accuracy": 0.840979814529419, + "num_tokens": 1766227.0, + "step": 245 + }, + { + "epoch": 0.25076452599388377, + "grad_norm": 5.335789203643799, + "learning_rate": 1e-06, + "loss": 0.381, + "mean_token_accuracy": 0.8485937118530273, + "num_tokens": 1773882.0, + "step": 246 + }, + { + "epoch": 0.25178389398572887, + "grad_norm": 5.599448204040527, + "learning_rate": 1e-06, + "loss": 0.4204, + "mean_token_accuracy": 0.8335893154144287, + "num_tokens": 1781330.0, + "step": 247 + }, + { + "epoch": 0.2528032619775739, + "grad_norm": 4.453589916229248, + "learning_rate": 1e-06, + "loss": 0.3743, + "mean_token_accuracy": 0.8491920828819275, + "num_tokens": 1789987.0, + "step": 248 + }, + { + "epoch": 0.25382262996941896, + "grad_norm": 6.619284629821777, + "learning_rate": 1e-06, + "loss": 0.3831, + "mean_token_accuracy": 0.8486117720603943, + "num_tokens": 1795881.0, + "step": 249 + }, + { + "epoch": 0.254841997961264, + "grad_norm": 6.908748626708984, + "learning_rate": 1e-06, + "loss": 0.4261, + "mean_token_accuracy": 0.8389949798583984, + "num_tokens": 1802386.0, + "step": 250 + }, + { + "epoch": 0.2558613659531091, + "grad_norm": 5.2094950675964355, + "learning_rate": 1e-06, + "loss": 0.3695, + "mean_token_accuracy": 0.8516823649406433, + "num_tokens": 1810348.0, + "step": 251 + }, + { + "epoch": 0.25688073394495414, + "grad_norm": 4.321964263916016, + "learning_rate": 1e-06, + "loss": 0.3502, + "mean_token_accuracy": 0.85009765625, + "num_tokens": 1819518.0, + "step": 252 + }, + { + "epoch": 0.2579001019367992, + "grad_norm": 5.596762180328369, + "learning_rate": 1e-06, + "loss": 0.3779, + "mean_token_accuracy": 0.8497060537338257, + "num_tokens": 1826620.0, + "step": 253 + }, + { + "epoch": 0.25891946992864423, + "grad_norm": 6.1245503425598145, + "learning_rate": 1e-06, + "loss": 0.4407, + "mean_token_accuracy": 0.828127384185791, + "num_tokens": 1833741.0, + "step": 254 + }, + { + "epoch": 0.2599388379204893, + "grad_norm": 4.61659574508667, + "learning_rate": 1e-06, + "loss": 0.3855, + "mean_token_accuracy": 0.8492653369903564, + "num_tokens": 1843297.0, + "step": 255 + }, + { + "epoch": 0.2609582059123344, + "grad_norm": 6.214087963104248, + "learning_rate": 1e-06, + "loss": 0.4084, + "mean_token_accuracy": 0.8338589072227478, + "num_tokens": 1850009.0, + "step": 256 + }, + { + "epoch": 0.2619775739041794, + "grad_norm": 4.631591320037842, + "learning_rate": 1e-06, + "loss": 0.3636, + "mean_token_accuracy": 0.8570421934127808, + "num_tokens": 1858811.0, + "step": 257 + }, + { + "epoch": 0.26299694189602446, + "grad_norm": 5.713013648986816, + "learning_rate": 1e-06, + "loss": 0.4057, + "mean_token_accuracy": 0.8326864242553711, + "num_tokens": 1866620.0, + "step": 258 + }, + { + "epoch": 0.2640163098878695, + "grad_norm": 8.05038833618164, + "learning_rate": 1e-06, + "loss": 0.438, + "mean_token_accuracy": 0.8323124647140503, + "num_tokens": 1873011.0, + "step": 259 + }, + { + "epoch": 0.2650356778797146, + "grad_norm": 6.0846967697143555, + "learning_rate": 1e-06, + "loss": 0.4169, + "mean_token_accuracy": 0.8360472917556763, + "num_tokens": 1879496.0, + "step": 260 + }, + { + "epoch": 0.26605504587155965, + "grad_norm": 5.926863193511963, + "learning_rate": 1e-06, + "loss": 0.4122, + "mean_token_accuracy": 0.8337969183921814, + "num_tokens": 1886654.0, + "step": 261 + }, + { + "epoch": 0.2670744138634047, + "grad_norm": 6.639545440673828, + "learning_rate": 1e-06, + "loss": 0.3985, + "mean_token_accuracy": 0.843428373336792, + "num_tokens": 1892547.0, + "step": 262 + }, + { + "epoch": 0.26809378185524974, + "grad_norm": 6.725264072418213, + "learning_rate": 1e-06, + "loss": 0.404, + "mean_token_accuracy": 0.838692843914032, + "num_tokens": 1898235.0, + "step": 263 + }, + { + "epoch": 0.2691131498470948, + "grad_norm": 6.294256210327148, + "learning_rate": 1e-06, + "loss": 0.3986, + "mean_token_accuracy": 0.8395344018936157, + "num_tokens": 1904437.0, + "step": 264 + }, + { + "epoch": 0.2701325178389399, + "grad_norm": 4.911297798156738, + "learning_rate": 1e-06, + "loss": 0.397, + "mean_token_accuracy": 0.8406201601028442, + "num_tokens": 1913180.0, + "step": 265 + }, + { + "epoch": 0.2711518858307849, + "grad_norm": 6.02631139755249, + "learning_rate": 1e-06, + "loss": 0.4199, + "mean_token_accuracy": 0.8352310657501221, + "num_tokens": 1919531.0, + "step": 266 + }, + { + "epoch": 0.27217125382262997, + "grad_norm": 4.940171718597412, + "learning_rate": 1e-06, + "loss": 0.3515, + "mean_token_accuracy": 0.8611208200454712, + "num_tokens": 1927288.0, + "step": 267 + }, + { + "epoch": 0.273190621814475, + "grad_norm": 6.115046501159668, + "learning_rate": 1e-06, + "loss": 0.4243, + "mean_token_accuracy": 0.8339431285858154, + "num_tokens": 1934349.0, + "step": 268 + }, + { + "epoch": 0.2742099898063201, + "grad_norm": 5.792862892150879, + "learning_rate": 1e-06, + "loss": 0.4322, + "mean_token_accuracy": 0.8314632177352905, + "num_tokens": 1941175.0, + "step": 269 + }, + { + "epoch": 0.27522935779816515, + "grad_norm": 5.714006423950195, + "learning_rate": 1e-06, + "loss": 0.4266, + "mean_token_accuracy": 0.8348331451416016, + "num_tokens": 1948127.0, + "step": 270 + }, + { + "epoch": 0.2762487257900102, + "grad_norm": 6.502542018890381, + "learning_rate": 1e-06, + "loss": 0.4642, + "mean_token_accuracy": 0.8170890808105469, + "num_tokens": 1954108.0, + "step": 271 + }, + { + "epoch": 0.27726809378185524, + "grad_norm": 5.975719451904297, + "learning_rate": 1e-06, + "loss": 0.4216, + "mean_token_accuracy": 0.8355295658111572, + "num_tokens": 1961170.0, + "step": 272 + }, + { + "epoch": 0.2782874617737003, + "grad_norm": 5.905372142791748, + "learning_rate": 1e-06, + "loss": 0.4291, + "mean_token_accuracy": 0.8316455483436584, + "num_tokens": 1968325.0, + "step": 273 + }, + { + "epoch": 0.2793068297655454, + "grad_norm": 5.527670860290527, + "learning_rate": 1e-06, + "loss": 0.43, + "mean_token_accuracy": 0.8209388852119446, + "num_tokens": 1976791.0, + "step": 274 + }, + { + "epoch": 0.2803261977573904, + "grad_norm": 5.598142147064209, + "learning_rate": 1e-06, + "loss": 0.3784, + "mean_token_accuracy": 0.8526148200035095, + "num_tokens": 1984504.0, + "step": 275 + }, + { + "epoch": 0.28134556574923547, + "grad_norm": 6.3025665283203125, + "learning_rate": 1e-06, + "loss": 0.4361, + "mean_token_accuracy": 0.8320814967155457, + "num_tokens": 1990844.0, + "step": 276 + }, + { + "epoch": 0.2823649337410805, + "grad_norm": 6.011943817138672, + "learning_rate": 1e-06, + "loss": 0.41, + "mean_token_accuracy": 0.8369401097297668, + "num_tokens": 1997648.0, + "step": 277 + }, + { + "epoch": 0.28338430173292556, + "grad_norm": 5.396106719970703, + "learning_rate": 1e-06, + "loss": 0.3792, + "mean_token_accuracy": 0.8455325365066528, + "num_tokens": 2004160.0, + "step": 278 + }, + { + "epoch": 0.28440366972477066, + "grad_norm": 5.179161071777344, + "learning_rate": 1e-06, + "loss": 0.4064, + "mean_token_accuracy": 0.8347492218017578, + "num_tokens": 2011795.0, + "step": 279 + }, + { + "epoch": 0.2854230377166157, + "grad_norm": 6.259247779846191, + "learning_rate": 1e-06, + "loss": 0.4225, + "mean_token_accuracy": 0.8332370519638062, + "num_tokens": 2018453.0, + "step": 280 + }, + { + "epoch": 0.28644240570846075, + "grad_norm": 5.252021789550781, + "learning_rate": 1e-06, + "loss": 0.397, + "mean_token_accuracy": 0.8465452194213867, + "num_tokens": 2026745.0, + "step": 281 + }, + { + "epoch": 0.2874617737003058, + "grad_norm": 7.309598445892334, + "learning_rate": 1e-06, + "loss": 0.409, + "mean_token_accuracy": 0.8426238894462585, + "num_tokens": 2033217.0, + "step": 282 + }, + { + "epoch": 0.2884811416921509, + "grad_norm": 6.324431896209717, + "learning_rate": 1e-06, + "loss": 0.4071, + "mean_token_accuracy": 0.8400503396987915, + "num_tokens": 2039990.0, + "step": 283 + }, + { + "epoch": 0.28950050968399593, + "grad_norm": 4.54123067855835, + "learning_rate": 1e-06, + "loss": 0.3561, + "mean_token_accuracy": 0.8598418235778809, + "num_tokens": 2048356.0, + "step": 284 + }, + { + "epoch": 0.290519877675841, + "grad_norm": 5.1969523429870605, + "learning_rate": 1e-06, + "loss": 0.4069, + "mean_token_accuracy": 0.8377974033355713, + "num_tokens": 2056320.0, + "step": 285 + }, + { + "epoch": 0.291539245667686, + "grad_norm": 7.289471626281738, + "learning_rate": 1e-06, + "loss": 0.411, + "mean_token_accuracy": 0.8333467245101929, + "num_tokens": 2062313.0, + "step": 286 + }, + { + "epoch": 0.29255861365953106, + "grad_norm": 5.638091564178467, + "learning_rate": 1e-06, + "loss": 0.39, + "mean_token_accuracy": 0.8488833904266357, + "num_tokens": 2069249.0, + "step": 287 + }, + { + "epoch": 0.29357798165137616, + "grad_norm": 6.369836807250977, + "learning_rate": 1e-06, + "loss": 0.4071, + "mean_token_accuracy": 0.8465316295623779, + "num_tokens": 2075456.0, + "step": 288 + }, + { + "epoch": 0.2945973496432212, + "grad_norm": 5.001073360443115, + "learning_rate": 1e-06, + "loss": 0.4236, + "mean_token_accuracy": 0.8317891359329224, + "num_tokens": 2084237.0, + "step": 289 + }, + { + "epoch": 0.29561671763506625, + "grad_norm": 5.176795482635498, + "learning_rate": 1e-06, + "loss": 0.3659, + "mean_token_accuracy": 0.8512234687805176, + "num_tokens": 2092230.0, + "step": 290 + }, + { + "epoch": 0.2966360856269113, + "grad_norm": 5.4565911293029785, + "learning_rate": 1e-06, + "loss": 0.4152, + "mean_token_accuracy": 0.8338848352432251, + "num_tokens": 2100056.0, + "step": 291 + }, + { + "epoch": 0.2976554536187564, + "grad_norm": 6.066525459289551, + "learning_rate": 1e-06, + "loss": 0.3946, + "mean_token_accuracy": 0.851630687713623, + "num_tokens": 2107531.0, + "step": 292 + }, + { + "epoch": 0.29867482161060144, + "grad_norm": 4.923036575317383, + "learning_rate": 1e-06, + "loss": 0.401, + "mean_token_accuracy": 0.8371875882148743, + "num_tokens": 2116315.0, + "step": 293 + }, + { + "epoch": 0.2996941896024465, + "grad_norm": 4.362541198730469, + "learning_rate": 1e-06, + "loss": 0.3477, + "mean_token_accuracy": 0.8519695997238159, + "num_tokens": 2125644.0, + "step": 294 + }, + { + "epoch": 0.3007135575942915, + "grad_norm": 5.361085414886475, + "learning_rate": 1e-06, + "loss": 0.448, + "mean_token_accuracy": 0.8258727788925171, + "num_tokens": 2133305.0, + "step": 295 + }, + { + "epoch": 0.30173292558613657, + "grad_norm": 4.970592975616455, + "learning_rate": 1e-06, + "loss": 0.3661, + "mean_token_accuracy": 0.852319598197937, + "num_tokens": 2141725.0, + "step": 296 + }, + { + "epoch": 0.30275229357798167, + "grad_norm": 5.682602882385254, + "learning_rate": 1e-06, + "loss": 0.4089, + "mean_token_accuracy": 0.8422542810440063, + "num_tokens": 2148357.0, + "step": 297 + }, + { + "epoch": 0.3037716615698267, + "grad_norm": 5.8285136222839355, + "learning_rate": 1e-06, + "loss": 0.4065, + "mean_token_accuracy": 0.8415609002113342, + "num_tokens": 2155123.0, + "step": 298 + }, + { + "epoch": 0.30479102956167176, + "grad_norm": 5.340397357940674, + "learning_rate": 1e-06, + "loss": 0.4023, + "mean_token_accuracy": 0.8362969756126404, + "num_tokens": 2162140.0, + "step": 299 + }, + { + "epoch": 0.3058103975535168, + "grad_norm": 5.832844257354736, + "learning_rate": 1e-06, + "loss": 0.4237, + "mean_token_accuracy": 0.8263219594955444, + "num_tokens": 2169209.0, + "step": 300 + }, + { + "epoch": 0.3068297655453619, + "grad_norm": 4.8182477951049805, + "learning_rate": 1e-06, + "loss": 0.3575, + "mean_token_accuracy": 0.8600375652313232, + "num_tokens": 2177264.0, + "step": 301 + }, + { + "epoch": 0.30784913353720694, + "grad_norm": 5.543737888336182, + "learning_rate": 1e-06, + "loss": 0.44, + "mean_token_accuracy": 0.8314424157142639, + "num_tokens": 2184820.0, + "step": 302 + }, + { + "epoch": 0.308868501529052, + "grad_norm": 5.846214771270752, + "learning_rate": 1e-06, + "loss": 0.3828, + "mean_token_accuracy": 0.8515350818634033, + "num_tokens": 2191508.0, + "step": 303 + }, + { + "epoch": 0.30988786952089703, + "grad_norm": 5.312794208526611, + "learning_rate": 1e-06, + "loss": 0.431, + "mean_token_accuracy": 0.8305671811103821, + "num_tokens": 2199534.0, + "step": 304 + }, + { + "epoch": 0.3109072375127421, + "grad_norm": 5.493511199951172, + "learning_rate": 1e-06, + "loss": 0.3857, + "mean_token_accuracy": 0.8496659994125366, + "num_tokens": 2207492.0, + "step": 305 + }, + { + "epoch": 0.3119266055045872, + "grad_norm": 5.951510429382324, + "learning_rate": 1e-06, + "loss": 0.3912, + "mean_token_accuracy": 0.8489542007446289, + "num_tokens": 2213702.0, + "step": 306 + }, + { + "epoch": 0.3129459734964322, + "grad_norm": 5.372223377227783, + "learning_rate": 1e-06, + "loss": 0.4003, + "mean_token_accuracy": 0.8430455923080444, + "num_tokens": 2221860.0, + "step": 307 + }, + { + "epoch": 0.31396534148827726, + "grad_norm": 5.759235858917236, + "learning_rate": 1e-06, + "loss": 0.4318, + "mean_token_accuracy": 0.8338718414306641, + "num_tokens": 2228903.0, + "step": 308 + }, + { + "epoch": 0.3149847094801223, + "grad_norm": 6.828544616699219, + "learning_rate": 1e-06, + "loss": 0.4234, + "mean_token_accuracy": 0.836904764175415, + "num_tokens": 2235782.0, + "step": 309 + }, + { + "epoch": 0.3160040774719674, + "grad_norm": 6.078234672546387, + "learning_rate": 1e-06, + "loss": 0.4282, + "mean_token_accuracy": 0.8419528007507324, + "num_tokens": 2243117.0, + "step": 310 + }, + { + "epoch": 0.31702344546381245, + "grad_norm": 4.470830917358398, + "learning_rate": 1e-06, + "loss": 0.3587, + "mean_token_accuracy": 0.861567497253418, + "num_tokens": 2252178.0, + "step": 311 + }, + { + "epoch": 0.3180428134556575, + "grad_norm": 5.4143877029418945, + "learning_rate": 1e-06, + "loss": 0.3885, + "mean_token_accuracy": 0.8468019366264343, + "num_tokens": 2260424.0, + "step": 312 + }, + { + "epoch": 0.31906218144750254, + "grad_norm": 6.0695648193359375, + "learning_rate": 1e-06, + "loss": 0.4462, + "mean_token_accuracy": 0.8295867443084717, + "num_tokens": 2267831.0, + "step": 313 + }, + { + "epoch": 0.3200815494393476, + "grad_norm": 5.5163984298706055, + "learning_rate": 1e-06, + "loss": 0.4374, + "mean_token_accuracy": 0.8351688385009766, + "num_tokens": 2276208.0, + "step": 314 + }, + { + "epoch": 0.3211009174311927, + "grad_norm": 5.64641809463501, + "learning_rate": 1e-06, + "loss": 0.4065, + "mean_token_accuracy": 0.8424655199050903, + "num_tokens": 2283744.0, + "step": 315 + }, + { + "epoch": 0.3221202854230377, + "grad_norm": 6.095568656921387, + "learning_rate": 1e-06, + "loss": 0.4129, + "mean_token_accuracy": 0.8429444432258606, + "num_tokens": 2289708.0, + "step": 316 + }, + { + "epoch": 0.32313965341488277, + "grad_norm": 6.4545135498046875, + "learning_rate": 1e-06, + "loss": 0.432, + "mean_token_accuracy": 0.8307980298995972, + "num_tokens": 2296083.0, + "step": 317 + }, + { + "epoch": 0.3241590214067278, + "grad_norm": 5.506949424743652, + "learning_rate": 1e-06, + "loss": 0.3749, + "mean_token_accuracy": 0.851380467414856, + "num_tokens": 2303335.0, + "step": 318 + }, + { + "epoch": 0.3251783893985729, + "grad_norm": 5.722652912139893, + "learning_rate": 1e-06, + "loss": 0.3697, + "mean_token_accuracy": 0.8517647981643677, + "num_tokens": 2310582.0, + "step": 319 + }, + { + "epoch": 0.32619775739041795, + "grad_norm": 5.4798431396484375, + "learning_rate": 1e-06, + "loss": 0.3849, + "mean_token_accuracy": 0.8519089221954346, + "num_tokens": 2319141.0, + "step": 320 + }, + { + "epoch": 0.327217125382263, + "grad_norm": 5.642482757568359, + "learning_rate": 1e-06, + "loss": 0.3967, + "mean_token_accuracy": 0.8418415784835815, + "num_tokens": 2326498.0, + "step": 321 + }, + { + "epoch": 0.32823649337410804, + "grad_norm": 5.366608619689941, + "learning_rate": 1e-06, + "loss": 0.4137, + "mean_token_accuracy": 0.8384379148483276, + "num_tokens": 2333301.0, + "step": 322 + }, + { + "epoch": 0.3292558613659531, + "grad_norm": 5.623032093048096, + "learning_rate": 1e-06, + "loss": 0.4031, + "mean_token_accuracy": 0.8453172445297241, + "num_tokens": 2340079.0, + "step": 323 + }, + { + "epoch": 0.3302752293577982, + "grad_norm": 5.79036808013916, + "learning_rate": 1e-06, + "loss": 0.4063, + "mean_token_accuracy": 0.8394635319709778, + "num_tokens": 2346884.0, + "step": 324 + }, + { + "epoch": 0.3312945973496432, + "grad_norm": 5.737825870513916, + "learning_rate": 1e-06, + "loss": 0.441, + "mean_token_accuracy": 0.8260766267776489, + "num_tokens": 2354643.0, + "step": 325 + }, + { + "epoch": 0.33231396534148827, + "grad_norm": 4.81677770614624, + "learning_rate": 1e-06, + "loss": 0.3537, + "mean_token_accuracy": 0.8578274250030518, + "num_tokens": 2363441.0, + "step": 326 + }, + { + "epoch": 0.3333333333333333, + "grad_norm": 6.079090595245361, + "learning_rate": 1e-06, + "loss": 0.3975, + "mean_token_accuracy": 0.8385471701622009, + "num_tokens": 2370009.0, + "step": 327 + }, + { + "epoch": 0.3343527013251784, + "grad_norm": 5.517184734344482, + "learning_rate": 1e-06, + "loss": 0.393, + "mean_token_accuracy": 0.8392930030822754, + "num_tokens": 2377175.0, + "step": 328 + }, + { + "epoch": 0.33537206931702346, + "grad_norm": 5.5860748291015625, + "learning_rate": 1e-06, + "loss": 0.3668, + "mean_token_accuracy": 0.857780396938324, + "num_tokens": 2384021.0, + "step": 329 + }, + { + "epoch": 0.3363914373088685, + "grad_norm": 6.530083656311035, + "learning_rate": 1e-06, + "loss": 0.3937, + "mean_token_accuracy": 0.84108567237854, + "num_tokens": 2389322.0, + "step": 330 + }, + { + "epoch": 0.33741080530071355, + "grad_norm": 4.89731502532959, + "learning_rate": 1e-06, + "loss": 0.3993, + "mean_token_accuracy": 0.8436719179153442, + "num_tokens": 2397615.0, + "step": 331 + }, + { + "epoch": 0.3384301732925586, + "grad_norm": 4.882175922393799, + "learning_rate": 1e-06, + "loss": 0.3519, + "mean_token_accuracy": 0.8449636101722717, + "num_tokens": 2404771.0, + "step": 332 + }, + { + "epoch": 0.3394495412844037, + "grad_norm": 6.048334121704102, + "learning_rate": 1e-06, + "loss": 0.3943, + "mean_token_accuracy": 0.8433369398117065, + "num_tokens": 2411285.0, + "step": 333 + }, + { + "epoch": 0.34046890927624873, + "grad_norm": 5.248860836029053, + "learning_rate": 1e-06, + "loss": 0.4034, + "mean_token_accuracy": 0.8429857492446899, + "num_tokens": 2418982.0, + "step": 334 + }, + { + "epoch": 0.3414882772680938, + "grad_norm": 5.2732014656066895, + "learning_rate": 1e-06, + "loss": 0.3719, + "mean_token_accuracy": 0.8421266674995422, + "num_tokens": 2425399.0, + "step": 335 + }, + { + "epoch": 0.3425076452599388, + "grad_norm": 5.186985492706299, + "learning_rate": 1e-06, + "loss": 0.3923, + "mean_token_accuracy": 0.8386759757995605, + "num_tokens": 2432171.0, + "step": 336 + }, + { + "epoch": 0.3435270132517839, + "grad_norm": 5.568289756774902, + "learning_rate": 1e-06, + "loss": 0.4097, + "mean_token_accuracy": 0.8379006385803223, + "num_tokens": 2438907.0, + "step": 337 + }, + { + "epoch": 0.34454638124362896, + "grad_norm": 5.8229217529296875, + "learning_rate": 1e-06, + "loss": 0.3652, + "mean_token_accuracy": 0.8485157489776611, + "num_tokens": 2445533.0, + "step": 338 + }, + { + "epoch": 0.345565749235474, + "grad_norm": 4.836280345916748, + "learning_rate": 1e-06, + "loss": 0.4009, + "mean_token_accuracy": 0.8431210517883301, + "num_tokens": 2453908.0, + "step": 339 + }, + { + "epoch": 0.34658511722731905, + "grad_norm": 5.795967102050781, + "learning_rate": 1e-06, + "loss": 0.3888, + "mean_token_accuracy": 0.8470596671104431, + "num_tokens": 2460346.0, + "step": 340 + }, + { + "epoch": 0.3476044852191641, + "grad_norm": 5.264500141143799, + "learning_rate": 1e-06, + "loss": 0.3728, + "mean_token_accuracy": 0.8519434928894043, + "num_tokens": 2467669.0, + "step": 341 + }, + { + "epoch": 0.3486238532110092, + "grad_norm": 5.852938652038574, + "learning_rate": 1e-06, + "loss": 0.378, + "mean_token_accuracy": 0.8450000286102295, + "num_tokens": 2474896.0, + "step": 342 + }, + { + "epoch": 0.34964322120285424, + "grad_norm": 5.376862049102783, + "learning_rate": 1e-06, + "loss": 0.3843, + "mean_token_accuracy": 0.8527026176452637, + "num_tokens": 2481892.0, + "step": 343 + }, + { + "epoch": 0.3506625891946993, + "grad_norm": 6.398754596710205, + "learning_rate": 1e-06, + "loss": 0.4294, + "mean_token_accuracy": 0.8269540071487427, + "num_tokens": 2487833.0, + "step": 344 + }, + { + "epoch": 0.3516819571865443, + "grad_norm": 5.344085693359375, + "learning_rate": 1e-06, + "loss": 0.3792, + "mean_token_accuracy": 0.8474357724189758, + "num_tokens": 2494800.0, + "step": 345 + }, + { + "epoch": 0.3527013251783894, + "grad_norm": 6.647225856781006, + "learning_rate": 1e-06, + "loss": 0.4262, + "mean_token_accuracy": 0.83240807056427, + "num_tokens": 2500638.0, + "step": 346 + }, + { + "epoch": 0.35372069317023447, + "grad_norm": 5.761654376983643, + "learning_rate": 1e-06, + "loss": 0.4164, + "mean_token_accuracy": 0.8381087779998779, + "num_tokens": 2507535.0, + "step": 347 + }, + { + "epoch": 0.3547400611620795, + "grad_norm": 5.581521987915039, + "learning_rate": 1e-06, + "loss": 0.3834, + "mean_token_accuracy": 0.8472845554351807, + "num_tokens": 2514792.0, + "step": 348 + }, + { + "epoch": 0.35575942915392456, + "grad_norm": 5.771222114562988, + "learning_rate": 1e-06, + "loss": 0.3819, + "mean_token_accuracy": 0.8451123237609863, + "num_tokens": 2521191.0, + "step": 349 + }, + { + "epoch": 0.3567787971457696, + "grad_norm": 5.415647983551025, + "learning_rate": 1e-06, + "loss": 0.4177, + "mean_token_accuracy": 0.8410168290138245, + "num_tokens": 2528861.0, + "step": 350 + }, + { + "epoch": 0.3577981651376147, + "grad_norm": 6.024875640869141, + "learning_rate": 1e-06, + "loss": 0.4097, + "mean_token_accuracy": 0.8421992063522339, + "num_tokens": 2535525.0, + "step": 351 + }, + { + "epoch": 0.35881753312945974, + "grad_norm": 5.502791404724121, + "learning_rate": 1e-06, + "loss": 0.4192, + "mean_token_accuracy": 0.8287185430526733, + "num_tokens": 2542650.0, + "step": 352 + }, + { + "epoch": 0.3598369011213048, + "grad_norm": 6.354938507080078, + "learning_rate": 1e-06, + "loss": 0.4088, + "mean_token_accuracy": 0.8425371646881104, + "num_tokens": 2549189.0, + "step": 353 + }, + { + "epoch": 0.36085626911314983, + "grad_norm": 6.161181449890137, + "learning_rate": 1e-06, + "loss": 0.4035, + "mean_token_accuracy": 0.8408796191215515, + "num_tokens": 2554905.0, + "step": 354 + }, + { + "epoch": 0.36187563710499493, + "grad_norm": 4.979153156280518, + "learning_rate": 1e-06, + "loss": 0.3993, + "mean_token_accuracy": 0.8339028358459473, + "num_tokens": 2563448.0, + "step": 355 + }, + { + "epoch": 0.36289500509684, + "grad_norm": 5.395542144775391, + "learning_rate": 1e-06, + "loss": 0.3734, + "mean_token_accuracy": 0.8442685604095459, + "num_tokens": 2571348.0, + "step": 356 + }, + { + "epoch": 0.363914373088685, + "grad_norm": 6.105888366699219, + "learning_rate": 1e-06, + "loss": 0.3886, + "mean_token_accuracy": 0.8538435697555542, + "num_tokens": 2578266.0, + "step": 357 + }, + { + "epoch": 0.36493374108053006, + "grad_norm": 5.514764308929443, + "learning_rate": 1e-06, + "loss": 0.3901, + "mean_token_accuracy": 0.8424912095069885, + "num_tokens": 2586034.0, + "step": 358 + }, + { + "epoch": 0.3659531090723751, + "grad_norm": 5.788231372833252, + "learning_rate": 1e-06, + "loss": 0.3836, + "mean_token_accuracy": 0.8478330373764038, + "num_tokens": 2592565.0, + "step": 359 + }, + { + "epoch": 0.3669724770642202, + "grad_norm": 6.113404273986816, + "learning_rate": 1e-06, + "loss": 0.3649, + "mean_token_accuracy": 0.8532695770263672, + "num_tokens": 2599146.0, + "step": 360 + }, + { + "epoch": 0.36799184505606525, + "grad_norm": 5.402470111846924, + "learning_rate": 1e-06, + "loss": 0.4052, + "mean_token_accuracy": 0.8384994268417358, + "num_tokens": 2606118.0, + "step": 361 + }, + { + "epoch": 0.3690112130479103, + "grad_norm": 5.40577507019043, + "learning_rate": 1e-06, + "loss": 0.3882, + "mean_token_accuracy": 0.8446071147918701, + "num_tokens": 2613880.0, + "step": 362 + }, + { + "epoch": 0.37003058103975534, + "grad_norm": 5.0479655265808105, + "learning_rate": 1e-06, + "loss": 0.3804, + "mean_token_accuracy": 0.8535957336425781, + "num_tokens": 2621987.0, + "step": 363 + }, + { + "epoch": 0.37104994903160043, + "grad_norm": 5.282904148101807, + "learning_rate": 1e-06, + "loss": 0.3922, + "mean_token_accuracy": 0.8473160862922668, + "num_tokens": 2629061.0, + "step": 364 + }, + { + "epoch": 0.3720693170234455, + "grad_norm": 5.837626934051514, + "learning_rate": 1e-06, + "loss": 0.4438, + "mean_token_accuracy": 0.8251329064369202, + "num_tokens": 2637218.0, + "step": 365 + }, + { + "epoch": 0.3730886850152905, + "grad_norm": 5.45843505859375, + "learning_rate": 1e-06, + "loss": 0.451, + "mean_token_accuracy": 0.8266828656196594, + "num_tokens": 2644736.0, + "step": 366 + }, + { + "epoch": 0.37410805300713557, + "grad_norm": 5.851597309112549, + "learning_rate": 1e-06, + "loss": 0.427, + "mean_token_accuracy": 0.8304749727249146, + "num_tokens": 2652276.0, + "step": 367 + }, + { + "epoch": 0.3751274209989806, + "grad_norm": 5.600654602050781, + "learning_rate": 1e-06, + "loss": 0.3821, + "mean_token_accuracy": 0.8501452207565308, + "num_tokens": 2659548.0, + "step": 368 + }, + { + "epoch": 0.3761467889908257, + "grad_norm": 5.841953277587891, + "learning_rate": 1e-06, + "loss": 0.4136, + "mean_token_accuracy": 0.8374667167663574, + "num_tokens": 2666599.0, + "step": 369 + }, + { + "epoch": 0.37716615698267075, + "grad_norm": 5.593482494354248, + "learning_rate": 1e-06, + "loss": 0.3874, + "mean_token_accuracy": 0.8415655493736267, + "num_tokens": 2673139.0, + "step": 370 + }, + { + "epoch": 0.3781855249745158, + "grad_norm": 5.45664644241333, + "learning_rate": 1e-06, + "loss": 0.4108, + "mean_token_accuracy": 0.8354474306106567, + "num_tokens": 2680791.0, + "step": 371 + }, + { + "epoch": 0.37920489296636084, + "grad_norm": 5.581776142120361, + "learning_rate": 1e-06, + "loss": 0.3938, + "mean_token_accuracy": 0.8468232154846191, + "num_tokens": 2687744.0, + "step": 372 + }, + { + "epoch": 0.38022426095820594, + "grad_norm": 5.292571067810059, + "learning_rate": 1e-06, + "loss": 0.4309, + "mean_token_accuracy": 0.8308091163635254, + "num_tokens": 2696396.0, + "step": 373 + }, + { + "epoch": 0.381243628950051, + "grad_norm": 4.050578594207764, + "learning_rate": 1e-06, + "loss": 0.3294, + "mean_token_accuracy": 0.8724040985107422, + "num_tokens": 2705944.0, + "step": 374 + }, + { + "epoch": 0.382262996941896, + "grad_norm": 5.7585062980651855, + "learning_rate": 1e-06, + "loss": 0.3879, + "mean_token_accuracy": 0.8456628322601318, + "num_tokens": 2713363.0, + "step": 375 + }, + { + "epoch": 0.38328236493374107, + "grad_norm": 6.01189661026001, + "learning_rate": 1e-06, + "loss": 0.4397, + "mean_token_accuracy": 0.8269146680831909, + "num_tokens": 2719637.0, + "step": 376 + }, + { + "epoch": 0.3843017329255861, + "grad_norm": 6.0041375160217285, + "learning_rate": 1e-06, + "loss": 0.4024, + "mean_token_accuracy": 0.8410335779190063, + "num_tokens": 2726711.0, + "step": 377 + }, + { + "epoch": 0.3853211009174312, + "grad_norm": 4.852994441986084, + "learning_rate": 1e-06, + "loss": 0.3734, + "mean_token_accuracy": 0.8551684617996216, + "num_tokens": 2735234.0, + "step": 378 + }, + { + "epoch": 0.38634046890927626, + "grad_norm": 5.830019474029541, + "learning_rate": 1e-06, + "loss": 0.3852, + "mean_token_accuracy": 0.8563306331634521, + "num_tokens": 2741705.0, + "step": 379 + }, + { + "epoch": 0.3873598369011213, + "grad_norm": 6.203460216522217, + "learning_rate": 1e-06, + "loss": 0.4233, + "mean_token_accuracy": 0.8333289623260498, + "num_tokens": 2748591.0, + "step": 380 + }, + { + "epoch": 0.38837920489296635, + "grad_norm": 6.159877777099609, + "learning_rate": 1e-06, + "loss": 0.3923, + "mean_token_accuracy": 0.8460375070571899, + "num_tokens": 2754242.0, + "step": 381 + }, + { + "epoch": 0.3893985728848114, + "grad_norm": 5.4370222091674805, + "learning_rate": 1e-06, + "loss": 0.3628, + "mean_token_accuracy": 0.8591421246528625, + "num_tokens": 2760711.0, + "step": 382 + }, + { + "epoch": 0.3904179408766565, + "grad_norm": 4.71657133102417, + "learning_rate": 1e-06, + "loss": 0.3506, + "mean_token_accuracy": 0.8495973348617554, + "num_tokens": 2768804.0, + "step": 383 + }, + { + "epoch": 0.39143730886850153, + "grad_norm": 5.51285457611084, + "learning_rate": 1e-06, + "loss": 0.3993, + "mean_token_accuracy": 0.8430302143096924, + "num_tokens": 2775777.0, + "step": 384 + }, + { + "epoch": 0.3924566768603466, + "grad_norm": 5.005456924438477, + "learning_rate": 1e-06, + "loss": 0.3868, + "mean_token_accuracy": 0.8480000495910645, + "num_tokens": 2783783.0, + "step": 385 + }, + { + "epoch": 0.3934760448521916, + "grad_norm": 6.486949920654297, + "learning_rate": 1e-06, + "loss": 0.4191, + "mean_token_accuracy": 0.8426445126533508, + "num_tokens": 2789489.0, + "step": 386 + }, + { + "epoch": 0.3944954128440367, + "grad_norm": 6.198330879211426, + "learning_rate": 1e-06, + "loss": 0.3961, + "mean_token_accuracy": 0.840366005897522, + "num_tokens": 2795467.0, + "step": 387 + }, + { + "epoch": 0.39551478083588176, + "grad_norm": 4.6131510734558105, + "learning_rate": 1e-06, + "loss": 0.3998, + "mean_token_accuracy": 0.8411953449249268, + "num_tokens": 2804520.0, + "step": 388 + }, + { + "epoch": 0.3965341488277268, + "grad_norm": 5.685761451721191, + "learning_rate": 1e-06, + "loss": 0.4127, + "mean_token_accuracy": 0.837745189666748, + "num_tokens": 2812758.0, + "step": 389 + }, + { + "epoch": 0.39755351681957185, + "grad_norm": 5.3956193923950195, + "learning_rate": 1e-06, + "loss": 0.4067, + "mean_token_accuracy": 0.8456135392189026, + "num_tokens": 2821012.0, + "step": 390 + }, + { + "epoch": 0.3985728848114169, + "grad_norm": 5.0795488357543945, + "learning_rate": 1e-06, + "loss": 0.4036, + "mean_token_accuracy": 0.8380163908004761, + "num_tokens": 2828885.0, + "step": 391 + }, + { + "epoch": 0.399592252803262, + "grad_norm": 5.710812091827393, + "learning_rate": 1e-06, + "loss": 0.3973, + "mean_token_accuracy": 0.8402731418609619, + "num_tokens": 2836016.0, + "step": 392 + }, + { + "epoch": 0.40061162079510704, + "grad_norm": 4.63301420211792, + "learning_rate": 1e-06, + "loss": 0.3642, + "mean_token_accuracy": 0.8502323031425476, + "num_tokens": 2843907.0, + "step": 393 + }, + { + "epoch": 0.4016309887869521, + "grad_norm": 5.723988056182861, + "learning_rate": 1e-06, + "loss": 0.4205, + "mean_token_accuracy": 0.8387054800987244, + "num_tokens": 2850518.0, + "step": 394 + }, + { + "epoch": 0.4026503567787971, + "grad_norm": 5.35150671005249, + "learning_rate": 1e-06, + "loss": 0.3555, + "mean_token_accuracy": 0.8630257844924927, + "num_tokens": 2857453.0, + "step": 395 + }, + { + "epoch": 0.4036697247706422, + "grad_norm": 4.89179801940918, + "learning_rate": 1e-06, + "loss": 0.3609, + "mean_token_accuracy": 0.8563772439956665, + "num_tokens": 2865560.0, + "step": 396 + }, + { + "epoch": 0.40468909276248727, + "grad_norm": 5.177602291107178, + "learning_rate": 1e-06, + "loss": 0.3988, + "mean_token_accuracy": 0.8445689678192139, + "num_tokens": 2873248.0, + "step": 397 + }, + { + "epoch": 0.4057084607543323, + "grad_norm": 6.338223457336426, + "learning_rate": 1e-06, + "loss": 0.4529, + "mean_token_accuracy": 0.8242887854576111, + "num_tokens": 2879246.0, + "step": 398 + }, + { + "epoch": 0.40672782874617736, + "grad_norm": 5.229898452758789, + "learning_rate": 1e-06, + "loss": 0.3665, + "mean_token_accuracy": 0.8510274291038513, + "num_tokens": 2886585.0, + "step": 399 + }, + { + "epoch": 0.4077471967380224, + "grad_norm": 6.027231216430664, + "learning_rate": 1e-06, + "loss": 0.3943, + "mean_token_accuracy": 0.8456231951713562, + "num_tokens": 2893145.0, + "step": 400 + }, + { + "epoch": 0.4087665647298675, + "grad_norm": 6.229056358337402, + "learning_rate": 1e-06, + "loss": 0.4249, + "mean_token_accuracy": 0.8318268060684204, + "num_tokens": 2899397.0, + "step": 401 + }, + { + "epoch": 0.40978593272171254, + "grad_norm": 5.749095916748047, + "learning_rate": 1e-06, + "loss": 0.4055, + "mean_token_accuracy": 0.8464140295982361, + "num_tokens": 2907045.0, + "step": 402 + }, + { + "epoch": 0.4108053007135576, + "grad_norm": 5.356473445892334, + "learning_rate": 1e-06, + "loss": 0.3732, + "mean_token_accuracy": 0.8449476957321167, + "num_tokens": 2914869.0, + "step": 403 + }, + { + "epoch": 0.41182466870540263, + "grad_norm": 5.330623149871826, + "learning_rate": 1e-06, + "loss": 0.4005, + "mean_token_accuracy": 0.8432578444480896, + "num_tokens": 2923628.0, + "step": 404 + }, + { + "epoch": 0.41284403669724773, + "grad_norm": 6.020007610321045, + "learning_rate": 1e-06, + "loss": 0.4014, + "mean_token_accuracy": 0.8443393111228943, + "num_tokens": 2929497.0, + "step": 405 + }, + { + "epoch": 0.4138634046890928, + "grad_norm": 6.041171550750732, + "learning_rate": 1e-06, + "loss": 0.4001, + "mean_token_accuracy": 0.8432130813598633, + "num_tokens": 2935971.0, + "step": 406 + }, + { + "epoch": 0.4148827726809378, + "grad_norm": 5.002862930297852, + "learning_rate": 1e-06, + "loss": 0.4049, + "mean_token_accuracy": 0.8418025970458984, + "num_tokens": 2945490.0, + "step": 407 + }, + { + "epoch": 0.41590214067278286, + "grad_norm": 5.27749490737915, + "learning_rate": 1e-06, + "loss": 0.3627, + "mean_token_accuracy": 0.8555173277854919, + "num_tokens": 2953319.0, + "step": 408 + }, + { + "epoch": 0.4169215086646279, + "grad_norm": 4.977192401885986, + "learning_rate": 1e-06, + "loss": 0.3655, + "mean_token_accuracy": 0.8521738052368164, + "num_tokens": 2960279.0, + "step": 409 + }, + { + "epoch": 0.417940876656473, + "grad_norm": 4.836373329162598, + "learning_rate": 1e-06, + "loss": 0.4111, + "mean_token_accuracy": 0.8373739719390869, + "num_tokens": 2969281.0, + "step": 410 + }, + { + "epoch": 0.41896024464831805, + "grad_norm": 5.842309474945068, + "learning_rate": 1e-06, + "loss": 0.4084, + "mean_token_accuracy": 0.8413679599761963, + "num_tokens": 2975642.0, + "step": 411 + }, + { + "epoch": 0.4199796126401631, + "grad_norm": 4.476048469543457, + "learning_rate": 1e-06, + "loss": 0.3416, + "mean_token_accuracy": 0.8582556247711182, + "num_tokens": 2984364.0, + "step": 412 + }, + { + "epoch": 0.42099898063200814, + "grad_norm": 5.51711368560791, + "learning_rate": 1e-06, + "loss": 0.3613, + "mean_token_accuracy": 0.8560575842857361, + "num_tokens": 2991398.0, + "step": 413 + }, + { + "epoch": 0.42201834862385323, + "grad_norm": 5.257907390594482, + "learning_rate": 1e-06, + "loss": 0.4052, + "mean_token_accuracy": 0.8356268405914307, + "num_tokens": 2998270.0, + "step": 414 + }, + { + "epoch": 0.4230377166156983, + "grad_norm": 5.389564514160156, + "learning_rate": 1e-06, + "loss": 0.4106, + "mean_token_accuracy": 0.8313031196594238, + "num_tokens": 3006053.0, + "step": 415 + }, + { + "epoch": 0.4240570846075433, + "grad_norm": 7.351229667663574, + "learning_rate": 1e-06, + "loss": 0.3913, + "mean_token_accuracy": 0.8472727537155151, + "num_tokens": 3011135.0, + "step": 416 + }, + { + "epoch": 0.42507645259938837, + "grad_norm": 5.601539134979248, + "learning_rate": 1e-06, + "loss": 0.3757, + "mean_token_accuracy": 0.8506463170051575, + "num_tokens": 3018085.0, + "step": 417 + }, + { + "epoch": 0.4260958205912334, + "grad_norm": 5.797730922698975, + "learning_rate": 1e-06, + "loss": 0.3897, + "mean_token_accuracy": 0.8500165939331055, + "num_tokens": 3024575.0, + "step": 418 + }, + { + "epoch": 0.4271151885830785, + "grad_norm": 4.962357997894287, + "learning_rate": 1e-06, + "loss": 0.3868, + "mean_token_accuracy": 0.8442628383636475, + "num_tokens": 3032726.0, + "step": 419 + }, + { + "epoch": 0.42813455657492355, + "grad_norm": 5.1378703117370605, + "learning_rate": 1e-06, + "loss": 0.4124, + "mean_token_accuracy": 0.8403026461601257, + "num_tokens": 3040772.0, + "step": 420 + }, + { + "epoch": 0.4291539245667686, + "grad_norm": 6.292715072631836, + "learning_rate": 1e-06, + "loss": 0.4148, + "mean_token_accuracy": 0.8367234468460083, + "num_tokens": 3047338.0, + "step": 421 + }, + { + "epoch": 0.43017329255861364, + "grad_norm": 4.978278160095215, + "learning_rate": 1e-06, + "loss": 0.3587, + "mean_token_accuracy": 0.8554505109786987, + "num_tokens": 3055061.0, + "step": 422 + }, + { + "epoch": 0.43119266055045874, + "grad_norm": 5.363531112670898, + "learning_rate": 1e-06, + "loss": 0.427, + "mean_token_accuracy": 0.8275718688964844, + "num_tokens": 3062893.0, + "step": 423 + }, + { + "epoch": 0.4322120285423038, + "grad_norm": 5.63649320602417, + "learning_rate": 1e-06, + "loss": 0.4031, + "mean_token_accuracy": 0.8393453359603882, + "num_tokens": 3069178.0, + "step": 424 + }, + { + "epoch": 0.4332313965341488, + "grad_norm": 5.830165863037109, + "learning_rate": 1e-06, + "loss": 0.3806, + "mean_token_accuracy": 0.8440372943878174, + "num_tokens": 3075417.0, + "step": 425 + }, + { + "epoch": 0.43425076452599387, + "grad_norm": 9.830938339233398, + "learning_rate": 1e-06, + "loss": 0.4191, + "mean_token_accuracy": 0.8362672328948975, + "num_tokens": 3083021.0, + "step": 426 + }, + { + "epoch": 0.4352701325178389, + "grad_norm": 7.429815769195557, + "learning_rate": 1e-06, + "loss": 0.3986, + "mean_token_accuracy": 0.8451384902000427, + "num_tokens": 3089322.0, + "step": 427 + }, + { + "epoch": 0.436289500509684, + "grad_norm": 5.486469268798828, + "learning_rate": 1e-06, + "loss": 0.3926, + "mean_token_accuracy": 0.8425695896148682, + "num_tokens": 3095503.0, + "step": 428 + }, + { + "epoch": 0.43730886850152906, + "grad_norm": 5.789192199707031, + "learning_rate": 1e-06, + "loss": 0.4122, + "mean_token_accuracy": 0.8355745077133179, + "num_tokens": 3102570.0, + "step": 429 + }, + { + "epoch": 0.4383282364933741, + "grad_norm": 6.295877933502197, + "learning_rate": 1e-06, + "loss": 0.4271, + "mean_token_accuracy": 0.8332194685935974, + "num_tokens": 3109072.0, + "step": 430 + }, + { + "epoch": 0.43934760448521915, + "grad_norm": 5.925631999969482, + "learning_rate": 1e-06, + "loss": 0.4184, + "mean_token_accuracy": 0.8344563245773315, + "num_tokens": 3115541.0, + "step": 431 + }, + { + "epoch": 0.44036697247706424, + "grad_norm": 6.861891746520996, + "learning_rate": 1e-06, + "loss": 0.42, + "mean_token_accuracy": 0.8370295763015747, + "num_tokens": 3120916.0, + "step": 432 + }, + { + "epoch": 0.4413863404689093, + "grad_norm": 5.764176845550537, + "learning_rate": 1e-06, + "loss": 0.3665, + "mean_token_accuracy": 0.8501923680305481, + "num_tokens": 3127346.0, + "step": 433 + }, + { + "epoch": 0.44240570846075433, + "grad_norm": 5.080046653747559, + "learning_rate": 1e-06, + "loss": 0.3796, + "mean_token_accuracy": 0.8523929119110107, + "num_tokens": 3135228.0, + "step": 434 + }, + { + "epoch": 0.4434250764525994, + "grad_norm": 5.491487979888916, + "learning_rate": 1e-06, + "loss": 0.3906, + "mean_token_accuracy": 0.8448950052261353, + "num_tokens": 3142726.0, + "step": 435 + }, + { + "epoch": 0.4444444444444444, + "grad_norm": 5.566355228424072, + "learning_rate": 1e-06, + "loss": 0.4571, + "mean_token_accuracy": 0.8221514225006104, + "num_tokens": 3150529.0, + "step": 436 + }, + { + "epoch": 0.4454638124362895, + "grad_norm": 5.4195427894592285, + "learning_rate": 1e-06, + "loss": 0.4083, + "mean_token_accuracy": 0.8429956436157227, + "num_tokens": 3157830.0, + "step": 437 + }, + { + "epoch": 0.44648318042813456, + "grad_norm": 4.770424842834473, + "learning_rate": 1e-06, + "loss": 0.3814, + "mean_token_accuracy": 0.851170539855957, + "num_tokens": 3165784.0, + "step": 438 + }, + { + "epoch": 0.4475025484199796, + "grad_norm": 5.526895046234131, + "learning_rate": 1e-06, + "loss": 0.405, + "mean_token_accuracy": 0.8465365171432495, + "num_tokens": 3173689.0, + "step": 439 + }, + { + "epoch": 0.44852191641182465, + "grad_norm": 5.591619491577148, + "learning_rate": 1e-06, + "loss": 0.3927, + "mean_token_accuracy": 0.8476258516311646, + "num_tokens": 3180540.0, + "step": 440 + }, + { + "epoch": 0.44954128440366975, + "grad_norm": 5.688458442687988, + "learning_rate": 1e-06, + "loss": 0.3925, + "mean_token_accuracy": 0.8516794443130493, + "num_tokens": 3187737.0, + "step": 441 + }, + { + "epoch": 0.4505606523955148, + "grad_norm": 5.454723358154297, + "learning_rate": 1e-06, + "loss": 0.359, + "mean_token_accuracy": 0.8542295098304749, + "num_tokens": 3195342.0, + "step": 442 + }, + { + "epoch": 0.45158002038735984, + "grad_norm": 7.028449535369873, + "learning_rate": 1e-06, + "loss": 0.4178, + "mean_token_accuracy": 0.8353744149208069, + "num_tokens": 3200696.0, + "step": 443 + }, + { + "epoch": 0.4525993883792049, + "grad_norm": 5.740809440612793, + "learning_rate": 1e-06, + "loss": 0.4047, + "mean_token_accuracy": 0.8413584232330322, + "num_tokens": 3207485.0, + "step": 444 + }, + { + "epoch": 0.4536187563710499, + "grad_norm": 4.152822017669678, + "learning_rate": 1e-06, + "loss": 0.3363, + "mean_token_accuracy": 0.8696759939193726, + "num_tokens": 3217342.0, + "step": 445 + }, + { + "epoch": 0.454638124362895, + "grad_norm": 5.0161967277526855, + "learning_rate": 1e-06, + "loss": 0.4074, + "mean_token_accuracy": 0.838234543800354, + "num_tokens": 3225284.0, + "step": 446 + }, + { + "epoch": 0.45565749235474007, + "grad_norm": 5.069014549255371, + "learning_rate": 1e-06, + "loss": 0.4095, + "mean_token_accuracy": 0.84492027759552, + "num_tokens": 3234422.0, + "step": 447 + }, + { + "epoch": 0.4566768603465851, + "grad_norm": 6.202249050140381, + "learning_rate": 1e-06, + "loss": 0.4209, + "mean_token_accuracy": 0.8364750146865845, + "num_tokens": 3240777.0, + "step": 448 + }, + { + "epoch": 0.45769622833843016, + "grad_norm": 4.748356819152832, + "learning_rate": 1e-06, + "loss": 0.3624, + "mean_token_accuracy": 0.8566171526908875, + "num_tokens": 3249675.0, + "step": 449 + }, + { + "epoch": 0.45871559633027525, + "grad_norm": 6.77650260925293, + "learning_rate": 1e-06, + "loss": 0.4764, + "mean_token_accuracy": 0.8190407752990723, + "num_tokens": 3256569.0, + "step": 450 + }, + { + "epoch": 0.4597349643221203, + "grad_norm": 5.077971935272217, + "learning_rate": 1e-06, + "loss": 0.376, + "mean_token_accuracy": 0.853737473487854, + "num_tokens": 3263829.0, + "step": 451 + }, + { + "epoch": 0.46075433231396534, + "grad_norm": 5.95535945892334, + "learning_rate": 1e-06, + "loss": 0.4101, + "mean_token_accuracy": 0.8340104818344116, + "num_tokens": 3270676.0, + "step": 452 + }, + { + "epoch": 0.4617737003058104, + "grad_norm": 6.145422458648682, + "learning_rate": 1e-06, + "loss": 0.4211, + "mean_token_accuracy": 0.8298593759536743, + "num_tokens": 3276970.0, + "step": 453 + }, + { + "epoch": 0.46279306829765543, + "grad_norm": 4.675235748291016, + "learning_rate": 1e-06, + "loss": 0.3733, + "mean_token_accuracy": 0.8506253361701965, + "num_tokens": 3286143.0, + "step": 454 + }, + { + "epoch": 0.46381243628950053, + "grad_norm": 5.428065776824951, + "learning_rate": 1e-06, + "loss": 0.4514, + "mean_token_accuracy": 0.820807695388794, + "num_tokens": 3293513.0, + "step": 455 + }, + { + "epoch": 0.4648318042813456, + "grad_norm": 6.461978435516357, + "learning_rate": 1e-06, + "loss": 0.4294, + "mean_token_accuracy": 0.8283825516700745, + "num_tokens": 3300746.0, + "step": 456 + }, + { + "epoch": 0.4658511722731906, + "grad_norm": 6.060362339019775, + "learning_rate": 1e-06, + "loss": 0.404, + "mean_token_accuracy": 0.840502917766571, + "num_tokens": 3306918.0, + "step": 457 + }, + { + "epoch": 0.46687054026503566, + "grad_norm": 5.521754264831543, + "learning_rate": 1e-06, + "loss": 0.395, + "mean_token_accuracy": 0.8395698070526123, + "num_tokens": 3313687.0, + "step": 458 + }, + { + "epoch": 0.46788990825688076, + "grad_norm": 5.671180248260498, + "learning_rate": 1e-06, + "loss": 0.3946, + "mean_token_accuracy": 0.8387681245803833, + "num_tokens": 3319657.0, + "step": 459 + }, + { + "epoch": 0.4689092762487258, + "grad_norm": 4.424076080322266, + "learning_rate": 1e-06, + "loss": 0.3808, + "mean_token_accuracy": 0.8526788949966431, + "num_tokens": 3328125.0, + "step": 460 + }, + { + "epoch": 0.46992864424057085, + "grad_norm": 5.796677589416504, + "learning_rate": 1e-06, + "loss": 0.3974, + "mean_token_accuracy": 0.8494836688041687, + "num_tokens": 3335862.0, + "step": 461 + }, + { + "epoch": 0.4709480122324159, + "grad_norm": 4.950414657592773, + "learning_rate": 1e-06, + "loss": 0.3919, + "mean_token_accuracy": 0.8454345464706421, + "num_tokens": 3344337.0, + "step": 462 + }, + { + "epoch": 0.47196738022426094, + "grad_norm": 5.6323018074035645, + "learning_rate": 1e-06, + "loss": 0.4074, + "mean_token_accuracy": 0.8400999307632446, + "num_tokens": 3351464.0, + "step": 463 + }, + { + "epoch": 0.47298674821610603, + "grad_norm": 5.17052698135376, + "learning_rate": 1e-06, + "loss": 0.3971, + "mean_token_accuracy": 0.8431624174118042, + "num_tokens": 3359178.0, + "step": 464 + }, + { + "epoch": 0.4740061162079511, + "grad_norm": 5.006285667419434, + "learning_rate": 1e-06, + "loss": 0.3957, + "mean_token_accuracy": 0.8377798199653625, + "num_tokens": 3367661.0, + "step": 465 + }, + { + "epoch": 0.4750254841997961, + "grad_norm": 5.12393856048584, + "learning_rate": 1e-06, + "loss": 0.3624, + "mean_token_accuracy": 0.8459917902946472, + "num_tokens": 3374749.0, + "step": 466 + }, + { + "epoch": 0.47604485219164117, + "grad_norm": 5.651128768920898, + "learning_rate": 1e-06, + "loss": 0.4202, + "mean_token_accuracy": 0.8366137742996216, + "num_tokens": 3382504.0, + "step": 467 + }, + { + "epoch": 0.47706422018348627, + "grad_norm": 6.672638893127441, + "learning_rate": 1e-06, + "loss": 0.4088, + "mean_token_accuracy": 0.8306671977043152, + "num_tokens": 3389114.0, + "step": 468 + }, + { + "epoch": 0.4780835881753313, + "grad_norm": 5.228642463684082, + "learning_rate": 1e-06, + "loss": 0.3603, + "mean_token_accuracy": 0.8565065860748291, + "num_tokens": 3396641.0, + "step": 469 + }, + { + "epoch": 0.47910295616717635, + "grad_norm": 5.930887222290039, + "learning_rate": 1e-06, + "loss": 0.4551, + "mean_token_accuracy": 0.820287823677063, + "num_tokens": 3403447.0, + "step": 470 + }, + { + "epoch": 0.4801223241590214, + "grad_norm": 4.966257095336914, + "learning_rate": 1e-06, + "loss": 0.3727, + "mean_token_accuracy": 0.8481652736663818, + "num_tokens": 3410744.0, + "step": 471 + }, + { + "epoch": 0.48114169215086644, + "grad_norm": 5.587610244750977, + "learning_rate": 1e-06, + "loss": 0.3971, + "mean_token_accuracy": 0.8425555229187012, + "num_tokens": 3418370.0, + "step": 472 + }, + { + "epoch": 0.48216106014271154, + "grad_norm": 4.64617395401001, + "learning_rate": 1e-06, + "loss": 0.3347, + "mean_token_accuracy": 0.862804114818573, + "num_tokens": 3426678.0, + "step": 473 + }, + { + "epoch": 0.4831804281345566, + "grad_norm": 6.230306148529053, + "learning_rate": 1e-06, + "loss": 0.3872, + "mean_token_accuracy": 0.8344675302505493, + "num_tokens": 3433327.0, + "step": 474 + }, + { + "epoch": 0.4841997961264016, + "grad_norm": 5.779729843139648, + "learning_rate": 1e-06, + "loss": 0.3887, + "mean_token_accuracy": 0.8475651144981384, + "num_tokens": 3439965.0, + "step": 475 + }, + { + "epoch": 0.48521916411824667, + "grad_norm": 5.184952735900879, + "learning_rate": 1e-06, + "loss": 0.3755, + "mean_token_accuracy": 0.855610191822052, + "num_tokens": 3448542.0, + "step": 476 + }, + { + "epoch": 0.48623853211009177, + "grad_norm": 5.547271251678467, + "learning_rate": 1e-06, + "loss": 0.3884, + "mean_token_accuracy": 0.8486493825912476, + "num_tokens": 3454818.0, + "step": 477 + }, + { + "epoch": 0.4872579001019368, + "grad_norm": 6.045424938201904, + "learning_rate": 1e-06, + "loss": 0.3753, + "mean_token_accuracy": 0.8509007096290588, + "num_tokens": 3460931.0, + "step": 478 + }, + { + "epoch": 0.48827726809378186, + "grad_norm": 5.539500713348389, + "learning_rate": 1e-06, + "loss": 0.4243, + "mean_token_accuracy": 0.8351383805274963, + "num_tokens": 3467945.0, + "step": 479 + }, + { + "epoch": 0.4892966360856269, + "grad_norm": 5.4876556396484375, + "learning_rate": 1e-06, + "loss": 0.4018, + "mean_token_accuracy": 0.8425442576408386, + "num_tokens": 3474830.0, + "step": 480 + }, + { + "epoch": 0.49031600407747195, + "grad_norm": 4.411192417144775, + "learning_rate": 1e-06, + "loss": 0.384, + "mean_token_accuracy": 0.8409337997436523, + "num_tokens": 3484742.0, + "step": 481 + }, + { + "epoch": 0.49133537206931704, + "grad_norm": 5.010730743408203, + "learning_rate": 1e-06, + "loss": 0.4419, + "mean_token_accuracy": 0.8336844444274902, + "num_tokens": 3493181.0, + "step": 482 + }, + { + "epoch": 0.4923547400611621, + "grad_norm": 4.7124409675598145, + "learning_rate": 1e-06, + "loss": 0.3883, + "mean_token_accuracy": 0.8465100526809692, + "num_tokens": 3501787.0, + "step": 483 + }, + { + "epoch": 0.49337410805300713, + "grad_norm": 6.106990814208984, + "learning_rate": 1e-06, + "loss": 0.4136, + "mean_token_accuracy": 0.8403345346450806, + "num_tokens": 3508123.0, + "step": 484 + }, + { + "epoch": 0.4943934760448522, + "grad_norm": 5.810024738311768, + "learning_rate": 1e-06, + "loss": 0.4499, + "mean_token_accuracy": 0.8275320529937744, + "num_tokens": 3515404.0, + "step": 485 + }, + { + "epoch": 0.4954128440366973, + "grad_norm": 5.696934223175049, + "learning_rate": 1e-06, + "loss": 0.4251, + "mean_token_accuracy": 0.8268289566040039, + "num_tokens": 3522762.0, + "step": 486 + }, + { + "epoch": 0.4964322120285423, + "grad_norm": 4.913833141326904, + "learning_rate": 1e-06, + "loss": 0.3915, + "mean_token_accuracy": 0.8487698435783386, + "num_tokens": 3531432.0, + "step": 487 + }, + { + "epoch": 0.49745158002038736, + "grad_norm": 6.034150123596191, + "learning_rate": 1e-06, + "loss": 0.4091, + "mean_token_accuracy": 0.8365657329559326, + "num_tokens": 3537345.0, + "step": 488 + }, + { + "epoch": 0.4984709480122324, + "grad_norm": 4.501239776611328, + "learning_rate": 1e-06, + "loss": 0.4031, + "mean_token_accuracy": 0.8420386910438538, + "num_tokens": 3546457.0, + "step": 489 + }, + { + "epoch": 0.49949031600407745, + "grad_norm": 4.907109260559082, + "learning_rate": 1e-06, + "loss": 0.3842, + "mean_token_accuracy": 0.8483879566192627, + "num_tokens": 3554069.0, + "step": 490 + }, + { + "epoch": 0.5005096839959225, + "grad_norm": 5.160167694091797, + "learning_rate": 1e-06, + "loss": 0.4159, + "mean_token_accuracy": 0.8369800448417664, + "num_tokens": 3562097.0, + "step": 491 + }, + { + "epoch": 0.5015290519877675, + "grad_norm": 5.252260208129883, + "learning_rate": 1e-06, + "loss": 0.3986, + "mean_token_accuracy": 0.8417607545852661, + "num_tokens": 3568887.0, + "step": 492 + }, + { + "epoch": 0.5025484199796126, + "grad_norm": 6.170266151428223, + "learning_rate": 1e-06, + "loss": 0.3988, + "mean_token_accuracy": 0.8444005250930786, + "num_tokens": 3575775.0, + "step": 493 + }, + { + "epoch": 0.5035677879714577, + "grad_norm": 4.5727362632751465, + "learning_rate": 1e-06, + "loss": 0.3296, + "mean_token_accuracy": 0.8647627830505371, + "num_tokens": 3583173.0, + "step": 494 + }, + { + "epoch": 0.5045871559633027, + "grad_norm": 5.026159763336182, + "learning_rate": 1e-06, + "loss": 0.4054, + "mean_token_accuracy": 0.8435075283050537, + "num_tokens": 3591062.0, + "step": 495 + }, + { + "epoch": 0.5056065239551478, + "grad_norm": 5.916635036468506, + "learning_rate": 1e-06, + "loss": 0.3799, + "mean_token_accuracy": 0.8469188809394836, + "num_tokens": 3597304.0, + "step": 496 + }, + { + "epoch": 0.5066258919469928, + "grad_norm": 5.3606743812561035, + "learning_rate": 1e-06, + "loss": 0.4017, + "mean_token_accuracy": 0.8373492360115051, + "num_tokens": 3604923.0, + "step": 497 + }, + { + "epoch": 0.5076452599388379, + "grad_norm": 5.494800567626953, + "learning_rate": 1e-06, + "loss": 0.397, + "mean_token_accuracy": 0.8373090028762817, + "num_tokens": 3611362.0, + "step": 498 + }, + { + "epoch": 0.508664627930683, + "grad_norm": 5.876342296600342, + "learning_rate": 1e-06, + "loss": 0.3928, + "mean_token_accuracy": 0.8433545827865601, + "num_tokens": 3618469.0, + "step": 499 + }, + { + "epoch": 0.509683995922528, + "grad_norm": 5.120059013366699, + "learning_rate": 1e-06, + "loss": 0.3575, + "mean_token_accuracy": 0.859999418258667, + "num_tokens": 3626014.0, + "step": 500 + }, + { + "epoch": 0.5107033639143731, + "grad_norm": 6.594124794006348, + "learning_rate": 1e-06, + "loss": 0.4293, + "mean_token_accuracy": 0.8267165422439575, + "num_tokens": 3631625.0, + "step": 501 + }, + { + "epoch": 0.5117227319062182, + "grad_norm": 6.2937421798706055, + "learning_rate": 1e-06, + "loss": 0.4609, + "mean_token_accuracy": 0.8226903676986694, + "num_tokens": 3638041.0, + "step": 502 + }, + { + "epoch": 0.5127420998980632, + "grad_norm": 5.827455520629883, + "learning_rate": 1e-06, + "loss": 0.3973, + "mean_token_accuracy": 0.8414038419723511, + "num_tokens": 3644270.0, + "step": 503 + }, + { + "epoch": 0.5137614678899083, + "grad_norm": 6.60481595993042, + "learning_rate": 1e-06, + "loss": 0.4376, + "mean_token_accuracy": 0.838979184627533, + "num_tokens": 3649861.0, + "step": 504 + }, + { + "epoch": 0.5147808358817533, + "grad_norm": 5.489433765411377, + "learning_rate": 1e-06, + "loss": 0.393, + "mean_token_accuracy": 0.8459085822105408, + "num_tokens": 3656402.0, + "step": 505 + }, + { + "epoch": 0.5158002038735984, + "grad_norm": 5.273266792297363, + "learning_rate": 1e-06, + "loss": 0.3823, + "mean_token_accuracy": 0.8499217629432678, + "num_tokens": 3663956.0, + "step": 506 + }, + { + "epoch": 0.5168195718654435, + "grad_norm": 5.6249895095825195, + "learning_rate": 1e-06, + "loss": 0.4023, + "mean_token_accuracy": 0.8411983251571655, + "num_tokens": 3671975.0, + "step": 507 + }, + { + "epoch": 0.5178389398572885, + "grad_norm": 4.92648983001709, + "learning_rate": 1e-06, + "loss": 0.325, + "mean_token_accuracy": 0.8712482452392578, + "num_tokens": 3679506.0, + "step": 508 + }, + { + "epoch": 0.5188583078491336, + "grad_norm": 5.400071144104004, + "learning_rate": 1e-06, + "loss": 0.4048, + "mean_token_accuracy": 0.8419044017791748, + "num_tokens": 3686984.0, + "step": 509 + }, + { + "epoch": 0.5198776758409785, + "grad_norm": 5.723862171173096, + "learning_rate": 1e-06, + "loss": 0.4024, + "mean_token_accuracy": 0.8479397296905518, + "num_tokens": 3694843.0, + "step": 510 + }, + { + "epoch": 0.5208970438328236, + "grad_norm": 5.6839070320129395, + "learning_rate": 1e-06, + "loss": 0.4362, + "mean_token_accuracy": 0.8341890573501587, + "num_tokens": 3702429.0, + "step": 511 + }, + { + "epoch": 0.5219164118246687, + "grad_norm": 6.684300422668457, + "learning_rate": 1e-06, + "loss": 0.451, + "mean_token_accuracy": 0.8230240345001221, + "num_tokens": 3709061.0, + "step": 512 + }, + { + "epoch": 0.5229357798165137, + "grad_norm": 4.944110870361328, + "learning_rate": 1e-06, + "loss": 0.3751, + "mean_token_accuracy": 0.8463661670684814, + "num_tokens": 3716866.0, + "step": 513 + }, + { + "epoch": 0.5239551478083588, + "grad_norm": 5.419219970703125, + "learning_rate": 1e-06, + "loss": 0.3961, + "mean_token_accuracy": 0.8441585302352905, + "num_tokens": 3724625.0, + "step": 514 + }, + { + "epoch": 0.5249745158002038, + "grad_norm": 5.913327217102051, + "learning_rate": 1e-06, + "loss": 0.3941, + "mean_token_accuracy": 0.8416087627410889, + "num_tokens": 3730862.0, + "step": 515 + }, + { + "epoch": 0.5259938837920489, + "grad_norm": 5.448538780212402, + "learning_rate": 1e-06, + "loss": 0.4087, + "mean_token_accuracy": 0.8372958898544312, + "num_tokens": 3737729.0, + "step": 516 + }, + { + "epoch": 0.527013251783894, + "grad_norm": 5.280824184417725, + "learning_rate": 1e-06, + "loss": 0.3769, + "mean_token_accuracy": 0.8557437062263489, + "num_tokens": 3744620.0, + "step": 517 + }, + { + "epoch": 0.528032619775739, + "grad_norm": 5.579820156097412, + "learning_rate": 1e-06, + "loss": 0.388, + "mean_token_accuracy": 0.8415235877037048, + "num_tokens": 3753377.0, + "step": 518 + }, + { + "epoch": 0.5290519877675841, + "grad_norm": 5.703059673309326, + "learning_rate": 1e-06, + "loss": 0.3574, + "mean_token_accuracy": 0.8550187349319458, + "num_tokens": 3760263.0, + "step": 519 + }, + { + "epoch": 0.5300713557594292, + "grad_norm": 4.57135534286499, + "learning_rate": 1e-06, + "loss": 0.368, + "mean_token_accuracy": 0.8563207387924194, + "num_tokens": 3768715.0, + "step": 520 + }, + { + "epoch": 0.5310907237512742, + "grad_norm": 4.873245716094971, + "learning_rate": 1e-06, + "loss": 0.3866, + "mean_token_accuracy": 0.8453762531280518, + "num_tokens": 3777730.0, + "step": 521 + }, + { + "epoch": 0.5321100917431193, + "grad_norm": 6.182192802429199, + "learning_rate": 1e-06, + "loss": 0.4317, + "mean_token_accuracy": 0.8398605585098267, + "num_tokens": 3784461.0, + "step": 522 + }, + { + "epoch": 0.5331294597349643, + "grad_norm": 5.329368591308594, + "learning_rate": 1e-06, + "loss": 0.4084, + "mean_token_accuracy": 0.8342418074607849, + "num_tokens": 3792494.0, + "step": 523 + }, + { + "epoch": 0.5341488277268094, + "grad_norm": 5.731077671051025, + "learning_rate": 1e-06, + "loss": 0.4213, + "mean_token_accuracy": 0.8303678631782532, + "num_tokens": 3799350.0, + "step": 524 + }, + { + "epoch": 0.5351681957186545, + "grad_norm": 5.698396682739258, + "learning_rate": 1e-06, + "loss": 0.3956, + "mean_token_accuracy": 0.8442800045013428, + "num_tokens": 3806565.0, + "step": 525 + }, + { + "epoch": 0.5361875637104995, + "grad_norm": 6.147392749786377, + "learning_rate": 1e-06, + "loss": 0.3788, + "mean_token_accuracy": 0.8485590219497681, + "num_tokens": 3812467.0, + "step": 526 + }, + { + "epoch": 0.5372069317023446, + "grad_norm": 5.257273197174072, + "learning_rate": 1e-06, + "loss": 0.3716, + "mean_token_accuracy": 0.851940393447876, + "num_tokens": 3819404.0, + "step": 527 + }, + { + "epoch": 0.5382262996941896, + "grad_norm": 5.353507995605469, + "learning_rate": 1e-06, + "loss": 0.4173, + "mean_token_accuracy": 0.8298521637916565, + "num_tokens": 3826136.0, + "step": 528 + }, + { + "epoch": 0.5392456676860347, + "grad_norm": 5.326359272003174, + "learning_rate": 1e-06, + "loss": 0.3906, + "mean_token_accuracy": 0.8430888652801514, + "num_tokens": 3833220.0, + "step": 529 + }, + { + "epoch": 0.5402650356778798, + "grad_norm": 5.559825897216797, + "learning_rate": 1e-06, + "loss": 0.4005, + "mean_token_accuracy": 0.8404608964920044, + "num_tokens": 3840281.0, + "step": 530 + }, + { + "epoch": 0.5412844036697247, + "grad_norm": 6.46541166305542, + "learning_rate": 1e-06, + "loss": 0.3711, + "mean_token_accuracy": 0.850534200668335, + "num_tokens": 3845474.0, + "step": 531 + }, + { + "epoch": 0.5423037716615698, + "grad_norm": 5.870523929595947, + "learning_rate": 1e-06, + "loss": 0.411, + "mean_token_accuracy": 0.8319207429885864, + "num_tokens": 3852771.0, + "step": 532 + }, + { + "epoch": 0.5433231396534148, + "grad_norm": 6.0440673828125, + "learning_rate": 1e-06, + "loss": 0.4225, + "mean_token_accuracy": 0.8329952955245972, + "num_tokens": 3859591.0, + "step": 533 + }, + { + "epoch": 0.5443425076452599, + "grad_norm": 6.066277980804443, + "learning_rate": 1e-06, + "loss": 0.4222, + "mean_token_accuracy": 0.8286898136138916, + "num_tokens": 3866396.0, + "step": 534 + }, + { + "epoch": 0.545361875637105, + "grad_norm": 6.101741790771484, + "learning_rate": 1e-06, + "loss": 0.405, + "mean_token_accuracy": 0.8403414487838745, + "num_tokens": 3873000.0, + "step": 535 + }, + { + "epoch": 0.54638124362895, + "grad_norm": 4.862366199493408, + "learning_rate": 1e-06, + "loss": 0.3784, + "mean_token_accuracy": 0.8441982865333557, + "num_tokens": 3880776.0, + "step": 536 + }, + { + "epoch": 0.5474006116207951, + "grad_norm": 4.894218921661377, + "learning_rate": 1e-06, + "loss": 0.3991, + "mean_token_accuracy": 0.8386141061782837, + "num_tokens": 3888753.0, + "step": 537 + }, + { + "epoch": 0.5484199796126402, + "grad_norm": 4.62196683883667, + "learning_rate": 1e-06, + "loss": 0.3361, + "mean_token_accuracy": 0.8622297644615173, + "num_tokens": 3896089.0, + "step": 538 + }, + { + "epoch": 0.5494393476044852, + "grad_norm": 5.467430114746094, + "learning_rate": 1e-06, + "loss": 0.3899, + "mean_token_accuracy": 0.843148946762085, + "num_tokens": 3903110.0, + "step": 539 + }, + { + "epoch": 0.5504587155963303, + "grad_norm": 6.53594970703125, + "learning_rate": 1e-06, + "loss": 0.4116, + "mean_token_accuracy": 0.8338607549667358, + "num_tokens": 3909064.0, + "step": 540 + }, + { + "epoch": 0.5514780835881753, + "grad_norm": 5.201534748077393, + "learning_rate": 1e-06, + "loss": 0.3963, + "mean_token_accuracy": 0.8421118259429932, + "num_tokens": 3917035.0, + "step": 541 + }, + { + "epoch": 0.5524974515800204, + "grad_norm": 5.4246721267700195, + "learning_rate": 1e-06, + "loss": 0.4069, + "mean_token_accuracy": 0.8406596183776855, + "num_tokens": 3924720.0, + "step": 542 + }, + { + "epoch": 0.5535168195718655, + "grad_norm": 5.620701789855957, + "learning_rate": 1e-06, + "loss": 0.4072, + "mean_token_accuracy": 0.8355295658111572, + "num_tokens": 3931723.0, + "step": 543 + }, + { + "epoch": 0.5545361875637105, + "grad_norm": 5.562466144561768, + "learning_rate": 1e-06, + "loss": 0.3666, + "mean_token_accuracy": 0.8569539785385132, + "num_tokens": 3938911.0, + "step": 544 + }, + { + "epoch": 0.5555555555555556, + "grad_norm": 5.716331958770752, + "learning_rate": 1e-06, + "loss": 0.3989, + "mean_token_accuracy": 0.8398545980453491, + "num_tokens": 3946715.0, + "step": 545 + }, + { + "epoch": 0.5565749235474006, + "grad_norm": 5.460168361663818, + "learning_rate": 1e-06, + "loss": 0.3748, + "mean_token_accuracy": 0.8552985191345215, + "num_tokens": 3953645.0, + "step": 546 + }, + { + "epoch": 0.5575942915392457, + "grad_norm": 5.078230857849121, + "learning_rate": 1e-06, + "loss": 0.4041, + "mean_token_accuracy": 0.8429698944091797, + "num_tokens": 3960989.0, + "step": 547 + }, + { + "epoch": 0.5586136595310908, + "grad_norm": 5.005679130554199, + "learning_rate": 1e-06, + "loss": 0.4028, + "mean_token_accuracy": 0.8406301736831665, + "num_tokens": 3969860.0, + "step": 548 + }, + { + "epoch": 0.5596330275229358, + "grad_norm": 5.6393842697143555, + "learning_rate": 1e-06, + "loss": 0.4009, + "mean_token_accuracy": 0.8343037962913513, + "num_tokens": 3977356.0, + "step": 549 + }, + { + "epoch": 0.5606523955147809, + "grad_norm": 5.929019451141357, + "learning_rate": 1e-06, + "loss": 0.4053, + "mean_token_accuracy": 0.8375974893569946, + "num_tokens": 3984571.0, + "step": 550 + }, + { + "epoch": 0.5616717635066258, + "grad_norm": 5.591489791870117, + "learning_rate": 1e-06, + "loss": 0.4459, + "mean_token_accuracy": 0.8240095376968384, + "num_tokens": 3992359.0, + "step": 551 + }, + { + "epoch": 0.5626911314984709, + "grad_norm": 5.1364665031433105, + "learning_rate": 1e-06, + "loss": 0.3561, + "mean_token_accuracy": 0.8619619011878967, + "num_tokens": 4000517.0, + "step": 552 + }, + { + "epoch": 0.563710499490316, + "grad_norm": 5.455834865570068, + "learning_rate": 1e-06, + "loss": 0.4344, + "mean_token_accuracy": 0.8237200975418091, + "num_tokens": 4007770.0, + "step": 553 + }, + { + "epoch": 0.564729867482161, + "grad_norm": 5.180583953857422, + "learning_rate": 1e-06, + "loss": 0.3474, + "mean_token_accuracy": 0.8491863012313843, + "num_tokens": 4014859.0, + "step": 554 + }, + { + "epoch": 0.5657492354740061, + "grad_norm": 5.700765132904053, + "learning_rate": 1e-06, + "loss": 0.3798, + "mean_token_accuracy": 0.852209210395813, + "num_tokens": 4021473.0, + "step": 555 + }, + { + "epoch": 0.5667686034658511, + "grad_norm": 5.833667755126953, + "learning_rate": 1e-06, + "loss": 0.3632, + "mean_token_accuracy": 0.8562538027763367, + "num_tokens": 4028558.0, + "step": 556 + }, + { + "epoch": 0.5677879714576962, + "grad_norm": 5.870619773864746, + "learning_rate": 1e-06, + "loss": 0.4137, + "mean_token_accuracy": 0.844287633895874, + "num_tokens": 4035019.0, + "step": 557 + }, + { + "epoch": 0.5688073394495413, + "grad_norm": 5.383601665496826, + "learning_rate": 1e-06, + "loss": 0.4269, + "mean_token_accuracy": 0.8347702026367188, + "num_tokens": 4042607.0, + "step": 558 + }, + { + "epoch": 0.5698267074413863, + "grad_norm": 5.2430548667907715, + "learning_rate": 1e-06, + "loss": 0.407, + "mean_token_accuracy": 0.8398044109344482, + "num_tokens": 4049999.0, + "step": 559 + }, + { + "epoch": 0.5708460754332314, + "grad_norm": 5.609750270843506, + "learning_rate": 1e-06, + "loss": 0.4042, + "mean_token_accuracy": 0.8374696969985962, + "num_tokens": 4056159.0, + "step": 560 + }, + { + "epoch": 0.5718654434250765, + "grad_norm": 6.628569602966309, + "learning_rate": 1e-06, + "loss": 0.4234, + "mean_token_accuracy": 0.8357144594192505, + "num_tokens": 4061830.0, + "step": 561 + }, + { + "epoch": 0.5728848114169215, + "grad_norm": 5.9588775634765625, + "learning_rate": 1e-06, + "loss": 0.4359, + "mean_token_accuracy": 0.8276255130767822, + "num_tokens": 4068924.0, + "step": 562 + }, + { + "epoch": 0.5739041794087666, + "grad_norm": 5.832000255584717, + "learning_rate": 1e-06, + "loss": 0.3971, + "mean_token_accuracy": 0.8410657048225403, + "num_tokens": 4075953.0, + "step": 563 + }, + { + "epoch": 0.5749235474006116, + "grad_norm": 5.847659587860107, + "learning_rate": 1e-06, + "loss": 0.3835, + "mean_token_accuracy": 0.8447442054748535, + "num_tokens": 4083111.0, + "step": 564 + }, + { + "epoch": 0.5759429153924567, + "grad_norm": 6.790375709533691, + "learning_rate": 1e-06, + "loss": 0.3735, + "mean_token_accuracy": 0.8524745106697083, + "num_tokens": 4090449.0, + "step": 565 + }, + { + "epoch": 0.5769622833843018, + "grad_norm": 4.785898208618164, + "learning_rate": 1e-06, + "loss": 0.3946, + "mean_token_accuracy": 0.8533900380134583, + "num_tokens": 4098269.0, + "step": 566 + }, + { + "epoch": 0.5779816513761468, + "grad_norm": 5.924620151519775, + "learning_rate": 1e-06, + "loss": 0.4087, + "mean_token_accuracy": 0.8419901132583618, + "num_tokens": 4105504.0, + "step": 567 + }, + { + "epoch": 0.5790010193679919, + "grad_norm": 6.737812042236328, + "learning_rate": 1e-06, + "loss": 0.4338, + "mean_token_accuracy": 0.8257092833518982, + "num_tokens": 4110729.0, + "step": 568 + }, + { + "epoch": 0.5800203873598369, + "grad_norm": 5.636671543121338, + "learning_rate": 1e-06, + "loss": 0.3708, + "mean_token_accuracy": 0.8552433252334595, + "num_tokens": 4116854.0, + "step": 569 + }, + { + "epoch": 0.581039755351682, + "grad_norm": 4.879626274108887, + "learning_rate": 1e-06, + "loss": 0.3583, + "mean_token_accuracy": 0.8599926829338074, + "num_tokens": 4125596.0, + "step": 570 + }, + { + "epoch": 0.582059123343527, + "grad_norm": 5.561816215515137, + "learning_rate": 1e-06, + "loss": 0.4019, + "mean_token_accuracy": 0.8464774489402771, + "num_tokens": 4133469.0, + "step": 571 + }, + { + "epoch": 0.583078491335372, + "grad_norm": 5.553646087646484, + "learning_rate": 1e-06, + "loss": 0.4058, + "mean_token_accuracy": 0.8288795948028564, + "num_tokens": 4140256.0, + "step": 572 + }, + { + "epoch": 0.5840978593272171, + "grad_norm": 5.849698543548584, + "learning_rate": 1e-06, + "loss": 0.3909, + "mean_token_accuracy": 0.8483481407165527, + "num_tokens": 4147271.0, + "step": 573 + }, + { + "epoch": 0.5851172273190621, + "grad_norm": 5.986171245574951, + "learning_rate": 1e-06, + "loss": 0.428, + "mean_token_accuracy": 0.8355370163917542, + "num_tokens": 4154997.0, + "step": 574 + }, + { + "epoch": 0.5861365953109072, + "grad_norm": 5.069611072540283, + "learning_rate": 1e-06, + "loss": 0.3994, + "mean_token_accuracy": 0.8460514545440674, + "num_tokens": 4163165.0, + "step": 575 + }, + { + "epoch": 0.5871559633027523, + "grad_norm": 5.871618270874023, + "learning_rate": 1e-06, + "loss": 0.431, + "mean_token_accuracy": 0.8368635177612305, + "num_tokens": 4169863.0, + "step": 576 + }, + { + "epoch": 0.5881753312945973, + "grad_norm": 5.254565238952637, + "learning_rate": 1e-06, + "loss": 0.3827, + "mean_token_accuracy": 0.8454737663269043, + "num_tokens": 4178025.0, + "step": 577 + }, + { + "epoch": 0.5891946992864424, + "grad_norm": 5.914224624633789, + "learning_rate": 1e-06, + "loss": 0.3952, + "mean_token_accuracy": 0.8457037210464478, + "num_tokens": 4184301.0, + "step": 578 + }, + { + "epoch": 0.5902140672782875, + "grad_norm": 5.993251323699951, + "learning_rate": 1e-06, + "loss": 0.4291, + "mean_token_accuracy": 0.8331882357597351, + "num_tokens": 4191161.0, + "step": 579 + }, + { + "epoch": 0.5912334352701325, + "grad_norm": 5.9436726570129395, + "learning_rate": 1e-06, + "loss": 0.4153, + "mean_token_accuracy": 0.8389615416526794, + "num_tokens": 4197453.0, + "step": 580 + }, + { + "epoch": 0.5922528032619776, + "grad_norm": 5.003090858459473, + "learning_rate": 1e-06, + "loss": 0.37, + "mean_token_accuracy": 0.8510715961456299, + "num_tokens": 4205434.0, + "step": 581 + }, + { + "epoch": 0.5932721712538226, + "grad_norm": 5.875800132751465, + "learning_rate": 1e-06, + "loss": 0.4117, + "mean_token_accuracy": 0.8359978795051575, + "num_tokens": 4211835.0, + "step": 582 + }, + { + "epoch": 0.5942915392456677, + "grad_norm": 5.729613304138184, + "learning_rate": 1e-06, + "loss": 0.3929, + "mean_token_accuracy": 0.8450841307640076, + "num_tokens": 4218116.0, + "step": 583 + }, + { + "epoch": 0.5953109072375128, + "grad_norm": 6.131330490112305, + "learning_rate": 1e-06, + "loss": 0.4369, + "mean_token_accuracy": 0.833590030670166, + "num_tokens": 4223935.0, + "step": 584 + }, + { + "epoch": 0.5963302752293578, + "grad_norm": 5.15328311920166, + "learning_rate": 1e-06, + "loss": 0.376, + "mean_token_accuracy": 0.8484817743301392, + "num_tokens": 4231123.0, + "step": 585 + }, + { + "epoch": 0.5973496432212029, + "grad_norm": 6.628390789031982, + "learning_rate": 1e-06, + "loss": 0.4097, + "mean_token_accuracy": 0.8375191688537598, + "num_tokens": 4237122.0, + "step": 586 + }, + { + "epoch": 0.5983690112130479, + "grad_norm": 4.918139934539795, + "learning_rate": 1e-06, + "loss": 0.4141, + "mean_token_accuracy": 0.8397386074066162, + "num_tokens": 4246199.0, + "step": 587 + }, + { + "epoch": 0.599388379204893, + "grad_norm": 5.659829139709473, + "learning_rate": 1e-06, + "loss": 0.4018, + "mean_token_accuracy": 0.8403235077857971, + "num_tokens": 4252899.0, + "step": 588 + }, + { + "epoch": 0.6004077471967381, + "grad_norm": 5.460944652557373, + "learning_rate": 1e-06, + "loss": 0.4006, + "mean_token_accuracy": 0.835394024848938, + "num_tokens": 4260611.0, + "step": 589 + }, + { + "epoch": 0.601427115188583, + "grad_norm": 6.3940229415893555, + "learning_rate": 1e-06, + "loss": 0.4143, + "mean_token_accuracy": 0.8432638049125671, + "num_tokens": 4266479.0, + "step": 590 + }, + { + "epoch": 0.6024464831804281, + "grad_norm": 5.865124702453613, + "learning_rate": 1e-06, + "loss": 0.3496, + "mean_token_accuracy": 0.8571823835372925, + "num_tokens": 4272846.0, + "step": 591 + }, + { + "epoch": 0.6034658511722731, + "grad_norm": 5.6651387214660645, + "learning_rate": 1e-06, + "loss": 0.414, + "mean_token_accuracy": 0.8376376628875732, + "num_tokens": 4279889.0, + "step": 592 + }, + { + "epoch": 0.6044852191641182, + "grad_norm": 5.1185150146484375, + "learning_rate": 1e-06, + "loss": 0.4288, + "mean_token_accuracy": 0.8324509859085083, + "num_tokens": 4287953.0, + "step": 593 + }, + { + "epoch": 0.6055045871559633, + "grad_norm": 5.7017598152160645, + "learning_rate": 1e-06, + "loss": 0.415, + "mean_token_accuracy": 0.8366369009017944, + "num_tokens": 4295423.0, + "step": 594 + }, + { + "epoch": 0.6065239551478083, + "grad_norm": 5.930456638336182, + "learning_rate": 1e-06, + "loss": 0.3746, + "mean_token_accuracy": 0.8458586931228638, + "num_tokens": 4301570.0, + "step": 595 + }, + { + "epoch": 0.6075433231396534, + "grad_norm": 5.684777736663818, + "learning_rate": 1e-06, + "loss": 0.3821, + "mean_token_accuracy": 0.8506325483322144, + "num_tokens": 4308385.0, + "step": 596 + }, + { + "epoch": 0.6085626911314985, + "grad_norm": 4.772870063781738, + "learning_rate": 1e-06, + "loss": 0.3892, + "mean_token_accuracy": 0.844061017036438, + "num_tokens": 4317591.0, + "step": 597 + }, + { + "epoch": 0.6095820591233435, + "grad_norm": 5.409488677978516, + "learning_rate": 1e-06, + "loss": 0.3631, + "mean_token_accuracy": 0.8611012697219849, + "num_tokens": 4324955.0, + "step": 598 + }, + { + "epoch": 0.6106014271151886, + "grad_norm": 5.57073450088501, + "learning_rate": 1e-06, + "loss": 0.4162, + "mean_token_accuracy": 0.8342733383178711, + "num_tokens": 4332237.0, + "step": 599 + }, + { + "epoch": 0.6116207951070336, + "grad_norm": 5.419436931610107, + "learning_rate": 1e-06, + "loss": 0.4056, + "mean_token_accuracy": 0.8436402082443237, + "num_tokens": 4339513.0, + "step": 600 + }, + { + "epoch": 0.6126401630988787, + "grad_norm": 5.815826892852783, + "learning_rate": 1e-06, + "loss": 0.4034, + "mean_token_accuracy": 0.8396276831626892, + "num_tokens": 4346318.0, + "step": 601 + }, + { + "epoch": 0.6136595310907238, + "grad_norm": 4.398445129394531, + "learning_rate": 1e-06, + "loss": 0.3262, + "mean_token_accuracy": 0.8642606735229492, + "num_tokens": 4353747.0, + "step": 602 + }, + { + "epoch": 0.6146788990825688, + "grad_norm": 4.882534980773926, + "learning_rate": 1e-06, + "loss": 0.4006, + "mean_token_accuracy": 0.8383060097694397, + "num_tokens": 4361752.0, + "step": 603 + }, + { + "epoch": 0.6156982670744139, + "grad_norm": 4.872725486755371, + "learning_rate": 1e-06, + "loss": 0.3716, + "mean_token_accuracy": 0.8450945615768433, + "num_tokens": 4370022.0, + "step": 604 + }, + { + "epoch": 0.6167176350662589, + "grad_norm": 5.563065528869629, + "learning_rate": 1e-06, + "loss": 0.3833, + "mean_token_accuracy": 0.8459097146987915, + "num_tokens": 4376916.0, + "step": 605 + }, + { + "epoch": 0.617737003058104, + "grad_norm": 6.616767883300781, + "learning_rate": 1e-06, + "loss": 0.4255, + "mean_token_accuracy": 0.8238307237625122, + "num_tokens": 4382926.0, + "step": 606 + }, + { + "epoch": 0.6187563710499491, + "grad_norm": 4.834606647491455, + "learning_rate": 1e-06, + "loss": 0.3618, + "mean_token_accuracy": 0.8591049313545227, + "num_tokens": 4391191.0, + "step": 607 + }, + { + "epoch": 0.6197757390417941, + "grad_norm": 5.0129523277282715, + "learning_rate": 1e-06, + "loss": 0.4045, + "mean_token_accuracy": 0.8438847661018372, + "num_tokens": 4399081.0, + "step": 608 + }, + { + "epoch": 0.6207951070336392, + "grad_norm": 4.781810283660889, + "learning_rate": 1e-06, + "loss": 0.4059, + "mean_token_accuracy": 0.8411762118339539, + "num_tokens": 4407169.0, + "step": 609 + }, + { + "epoch": 0.6218144750254841, + "grad_norm": 5.063782215118408, + "learning_rate": 1e-06, + "loss": 0.3955, + "mean_token_accuracy": 0.8439310193061829, + "num_tokens": 4414838.0, + "step": 610 + }, + { + "epoch": 0.6228338430173292, + "grad_norm": 4.831152439117432, + "learning_rate": 1e-06, + "loss": 0.3912, + "mean_token_accuracy": 0.8464852571487427, + "num_tokens": 4423199.0, + "step": 611 + }, + { + "epoch": 0.6238532110091743, + "grad_norm": 6.210423946380615, + "learning_rate": 1e-06, + "loss": 0.395, + "mean_token_accuracy": 0.8450489044189453, + "num_tokens": 4428993.0, + "step": 612 + }, + { + "epoch": 0.6248725790010193, + "grad_norm": 5.880480766296387, + "learning_rate": 1e-06, + "loss": 0.4076, + "mean_token_accuracy": 0.8441241979598999, + "num_tokens": 4435264.0, + "step": 613 + }, + { + "epoch": 0.6258919469928644, + "grad_norm": 4.757497787475586, + "learning_rate": 1e-06, + "loss": 0.3794, + "mean_token_accuracy": 0.8514695167541504, + "num_tokens": 4443398.0, + "step": 614 + }, + { + "epoch": 0.6269113149847095, + "grad_norm": 5.541375160217285, + "learning_rate": 1e-06, + "loss": 0.3911, + "mean_token_accuracy": 0.8442391753196716, + "num_tokens": 4450138.0, + "step": 615 + }, + { + "epoch": 0.6279306829765545, + "grad_norm": 5.714560031890869, + "learning_rate": 1e-06, + "loss": 0.3897, + "mean_token_accuracy": 0.8429890275001526, + "num_tokens": 4456718.0, + "step": 616 + }, + { + "epoch": 0.6289500509683996, + "grad_norm": 5.579057216644287, + "learning_rate": 1e-06, + "loss": 0.4177, + "mean_token_accuracy": 0.840838611125946, + "num_tokens": 4464020.0, + "step": 617 + }, + { + "epoch": 0.6299694189602446, + "grad_norm": 5.11033821105957, + "learning_rate": 1e-06, + "loss": 0.341, + "mean_token_accuracy": 0.8635770678520203, + "num_tokens": 4470821.0, + "step": 618 + }, + { + "epoch": 0.6309887869520897, + "grad_norm": 6.097631931304932, + "learning_rate": 1e-06, + "loss": 0.4571, + "mean_token_accuracy": 0.8157694935798645, + "num_tokens": 4476835.0, + "step": 619 + }, + { + "epoch": 0.6320081549439348, + "grad_norm": 6.437216758728027, + "learning_rate": 1e-06, + "loss": 0.4453, + "mean_token_accuracy": 0.8205052614212036, + "num_tokens": 4483075.0, + "step": 620 + }, + { + "epoch": 0.6330275229357798, + "grad_norm": 5.014807224273682, + "learning_rate": 1e-06, + "loss": 0.3996, + "mean_token_accuracy": 0.8491454124450684, + "num_tokens": 4490663.0, + "step": 621 + }, + { + "epoch": 0.6340468909276249, + "grad_norm": 5.461190700531006, + "learning_rate": 1e-06, + "loss": 0.4, + "mean_token_accuracy": 0.8456059098243713, + "num_tokens": 4497573.0, + "step": 622 + }, + { + "epoch": 0.6350662589194699, + "grad_norm": 5.363071918487549, + "learning_rate": 1e-06, + "loss": 0.4133, + "mean_token_accuracy": 0.8322328329086304, + "num_tokens": 4504810.0, + "step": 623 + }, + { + "epoch": 0.636085626911315, + "grad_norm": 5.970504283905029, + "learning_rate": 1e-06, + "loss": 0.4002, + "mean_token_accuracy": 0.847058117389679, + "num_tokens": 4511404.0, + "step": 624 + }, + { + "epoch": 0.6371049949031601, + "grad_norm": 5.201910018920898, + "learning_rate": 1e-06, + "loss": 0.3685, + "mean_token_accuracy": 0.8517391085624695, + "num_tokens": 4518534.0, + "step": 625 + }, + { + "epoch": 0.6381243628950051, + "grad_norm": 5.534696102142334, + "learning_rate": 1e-06, + "loss": 0.4021, + "mean_token_accuracy": 0.8348854780197144, + "num_tokens": 4524766.0, + "step": 626 + }, + { + "epoch": 0.6391437308868502, + "grad_norm": 5.82359504699707, + "learning_rate": 1e-06, + "loss": 0.4198, + "mean_token_accuracy": 0.8372514247894287, + "num_tokens": 4530966.0, + "step": 627 + }, + { + "epoch": 0.6401630988786952, + "grad_norm": 4.970212936401367, + "learning_rate": 1e-06, + "loss": 0.4237, + "mean_token_accuracy": 0.8334447145462036, + "num_tokens": 4538769.0, + "step": 628 + }, + { + "epoch": 0.6411824668705403, + "grad_norm": 5.7281389236450195, + "learning_rate": 1e-06, + "loss": 0.3857, + "mean_token_accuracy": 0.8483232855796814, + "num_tokens": 4546080.0, + "step": 629 + }, + { + "epoch": 0.6422018348623854, + "grad_norm": 6.063386917114258, + "learning_rate": 1e-06, + "loss": 0.3737, + "mean_token_accuracy": 0.8585149049758911, + "num_tokens": 4552290.0, + "step": 630 + }, + { + "epoch": 0.6432212028542303, + "grad_norm": 6.255207061767578, + "learning_rate": 1e-06, + "loss": 0.4377, + "mean_token_accuracy": 0.8289065957069397, + "num_tokens": 4558713.0, + "step": 631 + }, + { + "epoch": 0.6442405708460754, + "grad_norm": 7.711328029632568, + "learning_rate": 1e-06, + "loss": 0.3649, + "mean_token_accuracy": 0.8569546341896057, + "num_tokens": 4565451.0, + "step": 632 + }, + { + "epoch": 0.6452599388379205, + "grad_norm": 5.902390480041504, + "learning_rate": 1e-06, + "loss": 0.4218, + "mean_token_accuracy": 0.8421436548233032, + "num_tokens": 4571832.0, + "step": 633 + }, + { + "epoch": 0.6462793068297655, + "grad_norm": 6.120721817016602, + "learning_rate": 1e-06, + "loss": 0.4207, + "mean_token_accuracy": 0.8352060317993164, + "num_tokens": 4578448.0, + "step": 634 + }, + { + "epoch": 0.6472986748216106, + "grad_norm": 6.220818042755127, + "learning_rate": 1e-06, + "loss": 0.3956, + "mean_token_accuracy": 0.8465399742126465, + "num_tokens": 4585194.0, + "step": 635 + }, + { + "epoch": 0.6483180428134556, + "grad_norm": 5.311255931854248, + "learning_rate": 1e-06, + "loss": 0.4153, + "mean_token_accuracy": 0.8360918760299683, + "num_tokens": 4593294.0, + "step": 636 + }, + { + "epoch": 0.6493374108053007, + "grad_norm": 5.193954944610596, + "learning_rate": 1e-06, + "loss": 0.3927, + "mean_token_accuracy": 0.8443034887313843, + "num_tokens": 4600362.0, + "step": 637 + }, + { + "epoch": 0.6503567787971458, + "grad_norm": 5.720885276794434, + "learning_rate": 1e-06, + "loss": 0.4163, + "mean_token_accuracy": 0.8338779211044312, + "num_tokens": 4607204.0, + "step": 638 + }, + { + "epoch": 0.6513761467889908, + "grad_norm": 4.430067539215088, + "learning_rate": 1e-06, + "loss": 0.3981, + "mean_token_accuracy": 0.8376588821411133, + "num_tokens": 4616457.0, + "step": 639 + }, + { + "epoch": 0.6523955147808359, + "grad_norm": 5.6224565505981445, + "learning_rate": 1e-06, + "loss": 0.4028, + "mean_token_accuracy": 0.8404260873794556, + "num_tokens": 4623645.0, + "step": 640 + }, + { + "epoch": 0.6534148827726809, + "grad_norm": 5.31981086730957, + "learning_rate": 1e-06, + "loss": 0.4302, + "mean_token_accuracy": 0.8293112516403198, + "num_tokens": 4631898.0, + "step": 641 + }, + { + "epoch": 0.654434250764526, + "grad_norm": 6.042684555053711, + "learning_rate": 1e-06, + "loss": 0.4113, + "mean_token_accuracy": 0.8364812135696411, + "num_tokens": 4637968.0, + "step": 642 + }, + { + "epoch": 0.6554536187563711, + "grad_norm": 5.010646343231201, + "learning_rate": 1e-06, + "loss": 0.3991, + "mean_token_accuracy": 0.8440191149711609, + "num_tokens": 4646302.0, + "step": 643 + }, + { + "epoch": 0.6564729867482161, + "grad_norm": 6.420506477355957, + "learning_rate": 1e-06, + "loss": 0.4311, + "mean_token_accuracy": 0.8290218710899353, + "num_tokens": 4652153.0, + "step": 644 + }, + { + "epoch": 0.6574923547400612, + "grad_norm": 6.276877403259277, + "learning_rate": 1e-06, + "loss": 0.4325, + "mean_token_accuracy": 0.8211249113082886, + "num_tokens": 4658027.0, + "step": 645 + }, + { + "epoch": 0.6585117227319062, + "grad_norm": 6.294401168823242, + "learning_rate": 1e-06, + "loss": 0.4302, + "mean_token_accuracy": 0.8313959240913391, + "num_tokens": 4664321.0, + "step": 646 + }, + { + "epoch": 0.6595310907237513, + "grad_norm": 5.332774639129639, + "learning_rate": 1e-06, + "loss": 0.3971, + "mean_token_accuracy": 0.8486981391906738, + "num_tokens": 4671346.0, + "step": 647 + }, + { + "epoch": 0.6605504587155964, + "grad_norm": 5.25991153717041, + "learning_rate": 1e-06, + "loss": 0.3642, + "mean_token_accuracy": 0.8578859567642212, + "num_tokens": 4679019.0, + "step": 648 + }, + { + "epoch": 0.6615698267074414, + "grad_norm": 5.046756267547607, + "learning_rate": 1e-06, + "loss": 0.3627, + "mean_token_accuracy": 0.8515605926513672, + "num_tokens": 4686637.0, + "step": 649 + }, + { + "epoch": 0.6625891946992865, + "grad_norm": 5.694122314453125, + "learning_rate": 1e-06, + "loss": 0.3782, + "mean_token_accuracy": 0.8481560945510864, + "num_tokens": 4693373.0, + "step": 650 + }, + { + "epoch": 0.6636085626911316, + "grad_norm": 5.343520164489746, + "learning_rate": 1e-06, + "loss": 0.4182, + "mean_token_accuracy": 0.8336284160614014, + "num_tokens": 4701963.0, + "step": 651 + }, + { + "epoch": 0.6646279306829765, + "grad_norm": 5.392237186431885, + "learning_rate": 1e-06, + "loss": 0.4047, + "mean_token_accuracy": 0.8346070051193237, + "num_tokens": 4709567.0, + "step": 652 + }, + { + "epoch": 0.6656472986748216, + "grad_norm": 6.587489604949951, + "learning_rate": 1e-06, + "loss": 0.4444, + "mean_token_accuracy": 0.8285250067710876, + "num_tokens": 4715963.0, + "step": 653 + }, + { + "epoch": 0.6666666666666666, + "grad_norm": 4.594203948974609, + "learning_rate": 1e-06, + "loss": 0.3396, + "mean_token_accuracy": 0.8604881763458252, + "num_tokens": 4724088.0, + "step": 654 + }, + { + "epoch": 0.6676860346585117, + "grad_norm": 5.110682487487793, + "learning_rate": 1e-06, + "loss": 0.3762, + "mean_token_accuracy": 0.8508889079093933, + "num_tokens": 4732212.0, + "step": 655 + }, + { + "epoch": 0.6687054026503568, + "grad_norm": 6.2202935218811035, + "learning_rate": 1e-06, + "loss": 0.434, + "mean_token_accuracy": 0.8324720859527588, + "num_tokens": 4739373.0, + "step": 656 + }, + { + "epoch": 0.6697247706422018, + "grad_norm": 4.60403299331665, + "learning_rate": 1e-06, + "loss": 0.3721, + "mean_token_accuracy": 0.8522201776504517, + "num_tokens": 4748124.0, + "step": 657 + }, + { + "epoch": 0.6707441386340469, + "grad_norm": 4.923756122589111, + "learning_rate": 1e-06, + "loss": 0.3778, + "mean_token_accuracy": 0.8465708494186401, + "num_tokens": 4756495.0, + "step": 658 + }, + { + "epoch": 0.6717635066258919, + "grad_norm": 5.99104642868042, + "learning_rate": 1e-06, + "loss": 0.438, + "mean_token_accuracy": 0.8303447961807251, + "num_tokens": 4763147.0, + "step": 659 + }, + { + "epoch": 0.672782874617737, + "grad_norm": 6.051051616668701, + "learning_rate": 1e-06, + "loss": 0.3825, + "mean_token_accuracy": 0.8494217395782471, + "num_tokens": 4769328.0, + "step": 660 + }, + { + "epoch": 0.6738022426095821, + "grad_norm": 5.557230472564697, + "learning_rate": 1e-06, + "loss": 0.4009, + "mean_token_accuracy": 0.8392518162727356, + "num_tokens": 4776597.0, + "step": 661 + }, + { + "epoch": 0.6748216106014271, + "grad_norm": 4.502664566040039, + "learning_rate": 1e-06, + "loss": 0.3772, + "mean_token_accuracy": 0.8527061939239502, + "num_tokens": 4786781.0, + "step": 662 + }, + { + "epoch": 0.6758409785932722, + "grad_norm": 6.106722831726074, + "learning_rate": 1e-06, + "loss": 0.4113, + "mean_token_accuracy": 0.8351379036903381, + "num_tokens": 4793284.0, + "step": 663 + }, + { + "epoch": 0.6768603465851172, + "grad_norm": 5.146297454833984, + "learning_rate": 1e-06, + "loss": 0.4058, + "mean_token_accuracy": 0.8444950580596924, + "num_tokens": 4799982.0, + "step": 664 + }, + { + "epoch": 0.6778797145769623, + "grad_norm": 5.133239269256592, + "learning_rate": 1e-06, + "loss": 0.3952, + "mean_token_accuracy": 0.8448889851570129, + "num_tokens": 4807458.0, + "step": 665 + }, + { + "epoch": 0.6788990825688074, + "grad_norm": 4.695289611816406, + "learning_rate": 1e-06, + "loss": 0.3929, + "mean_token_accuracy": 0.8409960269927979, + "num_tokens": 4816855.0, + "step": 666 + }, + { + "epoch": 0.6799184505606524, + "grad_norm": 5.531682968139648, + "learning_rate": 1e-06, + "loss": 0.3833, + "mean_token_accuracy": 0.8503329157829285, + "num_tokens": 4823036.0, + "step": 667 + }, + { + "epoch": 0.6809378185524975, + "grad_norm": 5.133326530456543, + "learning_rate": 1e-06, + "loss": 0.426, + "mean_token_accuracy": 0.835020124912262, + "num_tokens": 4831167.0, + "step": 668 + }, + { + "epoch": 0.6819571865443425, + "grad_norm": 5.855581283569336, + "learning_rate": 1e-06, + "loss": 0.4196, + "mean_token_accuracy": 0.8360694646835327, + "num_tokens": 4837163.0, + "step": 669 + }, + { + "epoch": 0.6829765545361876, + "grad_norm": 5.347293853759766, + "learning_rate": 1e-06, + "loss": 0.4066, + "mean_token_accuracy": 0.846117377281189, + "num_tokens": 4843907.0, + "step": 670 + }, + { + "epoch": 0.6839959225280327, + "grad_norm": 5.036080837249756, + "learning_rate": 1e-06, + "loss": 0.4111, + "mean_token_accuracy": 0.839458703994751, + "num_tokens": 4852099.0, + "step": 671 + }, + { + "epoch": 0.6850152905198776, + "grad_norm": 4.906449317932129, + "learning_rate": 1e-06, + "loss": 0.3831, + "mean_token_accuracy": 0.8494921922683716, + "num_tokens": 4860143.0, + "step": 672 + }, + { + "epoch": 0.6860346585117227, + "grad_norm": 5.271567344665527, + "learning_rate": 1e-06, + "loss": 0.4328, + "mean_token_accuracy": 0.8319408893585205, + "num_tokens": 4868200.0, + "step": 673 + }, + { + "epoch": 0.6870540265035678, + "grad_norm": 5.595365047454834, + "learning_rate": 1e-06, + "loss": 0.3808, + "mean_token_accuracy": 0.8456023931503296, + "num_tokens": 4875154.0, + "step": 674 + }, + { + "epoch": 0.6880733944954128, + "grad_norm": 5.591714382171631, + "learning_rate": 1e-06, + "loss": 0.3899, + "mean_token_accuracy": 0.843944251537323, + "num_tokens": 4881838.0, + "step": 675 + }, + { + "epoch": 0.6890927624872579, + "grad_norm": 5.495256423950195, + "learning_rate": 1e-06, + "loss": 0.4038, + "mean_token_accuracy": 0.8370711803436279, + "num_tokens": 4888978.0, + "step": 676 + }, + { + "epoch": 0.6901121304791029, + "grad_norm": 6.378079891204834, + "learning_rate": 1e-06, + "loss": 0.4035, + "mean_token_accuracy": 0.8416630625724792, + "num_tokens": 4895917.0, + "step": 677 + }, + { + "epoch": 0.691131498470948, + "grad_norm": 5.5352630615234375, + "learning_rate": 1e-06, + "loss": 0.4323, + "mean_token_accuracy": 0.8305766582489014, + "num_tokens": 4902765.0, + "step": 678 + }, + { + "epoch": 0.6921508664627931, + "grad_norm": 5.616952419281006, + "learning_rate": 1e-06, + "loss": 0.3979, + "mean_token_accuracy": 0.8467268943786621, + "num_tokens": 4909574.0, + "step": 679 + }, + { + "epoch": 0.6931702344546381, + "grad_norm": 6.702144145965576, + "learning_rate": 1e-06, + "loss": 0.3657, + "mean_token_accuracy": 0.8547569513320923, + "num_tokens": 4914468.0, + "step": 680 + }, + { + "epoch": 0.6941896024464832, + "grad_norm": 5.267213344573975, + "learning_rate": 1e-06, + "loss": 0.364, + "mean_token_accuracy": 0.8574105501174927, + "num_tokens": 4922275.0, + "step": 681 + }, + { + "epoch": 0.6952089704383282, + "grad_norm": 5.630412578582764, + "learning_rate": 1e-06, + "loss": 0.4023, + "mean_token_accuracy": 0.839464545249939, + "num_tokens": 4928662.0, + "step": 682 + }, + { + "epoch": 0.6962283384301733, + "grad_norm": 5.423763751983643, + "learning_rate": 1e-06, + "loss": 0.4277, + "mean_token_accuracy": 0.8321272134780884, + "num_tokens": 4935722.0, + "step": 683 + }, + { + "epoch": 0.6972477064220184, + "grad_norm": 5.933459281921387, + "learning_rate": 1e-06, + "loss": 0.3951, + "mean_token_accuracy": 0.8443712592124939, + "num_tokens": 4942261.0, + "step": 684 + }, + { + "epoch": 0.6982670744138634, + "grad_norm": 5.828713893890381, + "learning_rate": 1e-06, + "loss": 0.4145, + "mean_token_accuracy": 0.8320571184158325, + "num_tokens": 4948365.0, + "step": 685 + }, + { + "epoch": 0.6992864424057085, + "grad_norm": 5.752432346343994, + "learning_rate": 1e-06, + "loss": 0.4359, + "mean_token_accuracy": 0.822682797908783, + "num_tokens": 4955531.0, + "step": 686 + }, + { + "epoch": 0.7003058103975535, + "grad_norm": 4.261649131774902, + "learning_rate": 1e-06, + "loss": 0.3154, + "mean_token_accuracy": 0.8733741641044617, + "num_tokens": 4963931.0, + "step": 687 + }, + { + "epoch": 0.7013251783893986, + "grad_norm": 5.3911237716674805, + "learning_rate": 1e-06, + "loss": 0.4062, + "mean_token_accuracy": 0.8384361267089844, + "num_tokens": 4971817.0, + "step": 688 + }, + { + "epoch": 0.7023445463812437, + "grad_norm": 5.210681438446045, + "learning_rate": 1e-06, + "loss": 0.394, + "mean_token_accuracy": 0.8477672338485718, + "num_tokens": 4979719.0, + "step": 689 + }, + { + "epoch": 0.7033639143730887, + "grad_norm": 4.479341506958008, + "learning_rate": 1e-06, + "loss": 0.3756, + "mean_token_accuracy": 0.8509116172790527, + "num_tokens": 4988846.0, + "step": 690 + }, + { + "epoch": 0.7043832823649337, + "grad_norm": 6.445952415466309, + "learning_rate": 1e-06, + "loss": 0.447, + "mean_token_accuracy": 0.8147602081298828, + "num_tokens": 4994829.0, + "step": 691 + }, + { + "epoch": 0.7054026503567788, + "grad_norm": 5.890438079833984, + "learning_rate": 1e-06, + "loss": 0.3887, + "mean_token_accuracy": 0.844420313835144, + "num_tokens": 5001095.0, + "step": 692 + }, + { + "epoch": 0.7064220183486238, + "grad_norm": 5.635804176330566, + "learning_rate": 1e-06, + "loss": 0.4298, + "mean_token_accuracy": 0.837995707988739, + "num_tokens": 5007808.0, + "step": 693 + }, + { + "epoch": 0.7074413863404689, + "grad_norm": 5.529998302459717, + "learning_rate": 1e-06, + "loss": 0.3661, + "mean_token_accuracy": 0.8538627624511719, + "num_tokens": 5014016.0, + "step": 694 + }, + { + "epoch": 0.7084607543323139, + "grad_norm": 6.037661075592041, + "learning_rate": 1e-06, + "loss": 0.4379, + "mean_token_accuracy": 0.8286929130554199, + "num_tokens": 5020244.0, + "step": 695 + }, + { + "epoch": 0.709480122324159, + "grad_norm": 5.650250434875488, + "learning_rate": 1e-06, + "loss": 0.4049, + "mean_token_accuracy": 0.844770073890686, + "num_tokens": 5027483.0, + "step": 696 + }, + { + "epoch": 0.7104994903160041, + "grad_norm": 6.446862697601318, + "learning_rate": 1e-06, + "loss": 0.4104, + "mean_token_accuracy": 0.8321588039398193, + "num_tokens": 5033672.0, + "step": 697 + }, + { + "epoch": 0.7115188583078491, + "grad_norm": 5.379560470581055, + "learning_rate": 1e-06, + "loss": 0.3616, + "mean_token_accuracy": 0.8603780269622803, + "num_tokens": 5041221.0, + "step": 698 + }, + { + "epoch": 0.7125382262996942, + "grad_norm": 6.093654632568359, + "learning_rate": 1e-06, + "loss": 0.4224, + "mean_token_accuracy": 0.8371731042861938, + "num_tokens": 5047990.0, + "step": 699 + }, + { + "epoch": 0.7135575942915392, + "grad_norm": 5.603931427001953, + "learning_rate": 1e-06, + "loss": 0.3894, + "mean_token_accuracy": 0.842842698097229, + "num_tokens": 5054821.0, + "step": 700 + }, + { + "epoch": 0.7145769622833843, + "grad_norm": 5.275450229644775, + "learning_rate": 1e-06, + "loss": 0.4148, + "mean_token_accuracy": 0.8369609117507935, + "num_tokens": 5063153.0, + "step": 701 + }, + { + "epoch": 0.7155963302752294, + "grad_norm": 6.088174343109131, + "learning_rate": 1e-06, + "loss": 0.4237, + "mean_token_accuracy": 0.831531822681427, + "num_tokens": 5070297.0, + "step": 702 + }, + { + "epoch": 0.7166156982670744, + "grad_norm": 5.3647661209106445, + "learning_rate": 1e-06, + "loss": 0.3646, + "mean_token_accuracy": 0.8564911484718323, + "num_tokens": 5077997.0, + "step": 703 + }, + { + "epoch": 0.7176350662589195, + "grad_norm": 5.085275173187256, + "learning_rate": 1e-06, + "loss": 0.4081, + "mean_token_accuracy": 0.8436344861984253, + "num_tokens": 5085973.0, + "step": 704 + }, + { + "epoch": 0.7186544342507645, + "grad_norm": 5.317302703857422, + "learning_rate": 1e-06, + "loss": 0.4068, + "mean_token_accuracy": 0.8388984203338623, + "num_tokens": 5093987.0, + "step": 705 + }, + { + "epoch": 0.7196738022426096, + "grad_norm": 6.496004104614258, + "learning_rate": 1e-06, + "loss": 0.3975, + "mean_token_accuracy": 0.837910532951355, + "num_tokens": 5100099.0, + "step": 706 + }, + { + "epoch": 0.7206931702344547, + "grad_norm": 4.323646545410156, + "learning_rate": 1e-06, + "loss": 0.3606, + "mean_token_accuracy": 0.846282958984375, + "num_tokens": 5109465.0, + "step": 707 + }, + { + "epoch": 0.7217125382262997, + "grad_norm": 4.9358344078063965, + "learning_rate": 1e-06, + "loss": 0.3579, + "mean_token_accuracy": 0.8593147397041321, + "num_tokens": 5117547.0, + "step": 708 + }, + { + "epoch": 0.7227319062181448, + "grad_norm": 6.3661322593688965, + "learning_rate": 1e-06, + "loss": 0.4273, + "mean_token_accuracy": 0.8351686000823975, + "num_tokens": 5123107.0, + "step": 709 + }, + { + "epoch": 0.7237512742099899, + "grad_norm": 5.385876178741455, + "learning_rate": 1e-06, + "loss": 0.4098, + "mean_token_accuracy": 0.8376301527023315, + "num_tokens": 5130491.0, + "step": 710 + }, + { + "epoch": 0.7247706422018348, + "grad_norm": 5.334750652313232, + "learning_rate": 1e-06, + "loss": 0.396, + "mean_token_accuracy": 0.846774697303772, + "num_tokens": 5137490.0, + "step": 711 + }, + { + "epoch": 0.72579001019368, + "grad_norm": 6.0774006843566895, + "learning_rate": 1e-06, + "loss": 0.4226, + "mean_token_accuracy": 0.8313001394271851, + "num_tokens": 5144052.0, + "step": 712 + }, + { + "epoch": 0.7268093781855249, + "grad_norm": 5.411728858947754, + "learning_rate": 1e-06, + "loss": 0.4336, + "mean_token_accuracy": 0.8368760943412781, + "num_tokens": 5152026.0, + "step": 713 + }, + { + "epoch": 0.72782874617737, + "grad_norm": 5.007645130157471, + "learning_rate": 1e-06, + "loss": 0.4085, + "mean_token_accuracy": 0.8321250081062317, + "num_tokens": 5160085.0, + "step": 714 + }, + { + "epoch": 0.7288481141692151, + "grad_norm": 6.5201873779296875, + "learning_rate": 1e-06, + "loss": 0.3891, + "mean_token_accuracy": 0.8467147350311279, + "num_tokens": 5166814.0, + "step": 715 + }, + { + "epoch": 0.7298674821610601, + "grad_norm": 5.649572372436523, + "learning_rate": 1e-06, + "loss": 0.3941, + "mean_token_accuracy": 0.8530294895172119, + "num_tokens": 5173649.0, + "step": 716 + }, + { + "epoch": 0.7308868501529052, + "grad_norm": 5.376425266265869, + "learning_rate": 1e-06, + "loss": 0.4057, + "mean_token_accuracy": 0.8340373635292053, + "num_tokens": 5180661.0, + "step": 717 + }, + { + "epoch": 0.7319062181447502, + "grad_norm": 5.2692975997924805, + "learning_rate": 1e-06, + "loss": 0.3889, + "mean_token_accuracy": 0.8426460027694702, + "num_tokens": 5187788.0, + "step": 718 + }, + { + "epoch": 0.7329255861365953, + "grad_norm": 5.807580471038818, + "learning_rate": 1e-06, + "loss": 0.413, + "mean_token_accuracy": 0.8364399671554565, + "num_tokens": 5194483.0, + "step": 719 + }, + { + "epoch": 0.7339449541284404, + "grad_norm": 5.740751266479492, + "learning_rate": 1e-06, + "loss": 0.3912, + "mean_token_accuracy": 0.848637580871582, + "num_tokens": 5200886.0, + "step": 720 + }, + { + "epoch": 0.7349643221202854, + "grad_norm": 5.086346626281738, + "learning_rate": 1e-06, + "loss": 0.3877, + "mean_token_accuracy": 0.8462729454040527, + "num_tokens": 5208614.0, + "step": 721 + }, + { + "epoch": 0.7359836901121305, + "grad_norm": 5.202008247375488, + "learning_rate": 1e-06, + "loss": 0.3886, + "mean_token_accuracy": 0.8445619344711304, + "num_tokens": 5215574.0, + "step": 722 + }, + { + "epoch": 0.7370030581039755, + "grad_norm": 4.627859115600586, + "learning_rate": 1e-06, + "loss": 0.3736, + "mean_token_accuracy": 0.8526531457901001, + "num_tokens": 5223915.0, + "step": 723 + }, + { + "epoch": 0.7380224260958206, + "grad_norm": 5.208531856536865, + "learning_rate": 1e-06, + "loss": 0.3659, + "mean_token_accuracy": 0.8505218029022217, + "num_tokens": 5231388.0, + "step": 724 + }, + { + "epoch": 0.7390417940876657, + "grad_norm": 5.213199615478516, + "learning_rate": 1e-06, + "loss": 0.3978, + "mean_token_accuracy": 0.8487865924835205, + "num_tokens": 5239130.0, + "step": 725 + }, + { + "epoch": 0.7400611620795107, + "grad_norm": 5.2445902824401855, + "learning_rate": 1e-06, + "loss": 0.3747, + "mean_token_accuracy": 0.8541593551635742, + "num_tokens": 5247125.0, + "step": 726 + }, + { + "epoch": 0.7410805300713558, + "grad_norm": 5.934176445007324, + "learning_rate": 1e-06, + "loss": 0.3587, + "mean_token_accuracy": 0.855475664138794, + "num_tokens": 5253852.0, + "step": 727 + }, + { + "epoch": 0.7420998980632009, + "grad_norm": 5.529010772705078, + "learning_rate": 1e-06, + "loss": 0.4344, + "mean_token_accuracy": 0.8296109437942505, + "num_tokens": 5260638.0, + "step": 728 + }, + { + "epoch": 0.7431192660550459, + "grad_norm": 5.936758995056152, + "learning_rate": 1e-06, + "loss": 0.4073, + "mean_token_accuracy": 0.8424457311630249, + "num_tokens": 5267956.0, + "step": 729 + }, + { + "epoch": 0.744138634046891, + "grad_norm": 5.598800182342529, + "learning_rate": 1e-06, + "loss": 0.4053, + "mean_token_accuracy": 0.8334553837776184, + "num_tokens": 5275263.0, + "step": 730 + }, + { + "epoch": 0.745158002038736, + "grad_norm": 5.535443305969238, + "learning_rate": 1e-06, + "loss": 0.4119, + "mean_token_accuracy": 0.8354128003120422, + "num_tokens": 5282708.0, + "step": 731 + }, + { + "epoch": 0.746177370030581, + "grad_norm": 5.596563339233398, + "learning_rate": 1e-06, + "loss": 0.4227, + "mean_token_accuracy": 0.8359018564224243, + "num_tokens": 5289194.0, + "step": 732 + }, + { + "epoch": 0.7471967380224261, + "grad_norm": 5.5220947265625, + "learning_rate": 1e-06, + "loss": 0.4039, + "mean_token_accuracy": 0.841588020324707, + "num_tokens": 5295871.0, + "step": 733 + }, + { + "epoch": 0.7482161060142711, + "grad_norm": 5.4459052085876465, + "learning_rate": 1e-06, + "loss": 0.3813, + "mean_token_accuracy": 0.8486200571060181, + "num_tokens": 5303409.0, + "step": 734 + }, + { + "epoch": 0.7492354740061162, + "grad_norm": 5.015290260314941, + "learning_rate": 1e-06, + "loss": 0.4014, + "mean_token_accuracy": 0.8455119132995605, + "num_tokens": 5311301.0, + "step": 735 + }, + { + "epoch": 0.7502548419979612, + "grad_norm": 4.6090989112854, + "learning_rate": 1e-06, + "loss": 0.3783, + "mean_token_accuracy": 0.850684404373169, + "num_tokens": 5320672.0, + "step": 736 + }, + { + "epoch": 0.7512742099898063, + "grad_norm": 5.28950834274292, + "learning_rate": 1e-06, + "loss": 0.43, + "mean_token_accuracy": 0.8316044211387634, + "num_tokens": 5327508.0, + "step": 737 + }, + { + "epoch": 0.7522935779816514, + "grad_norm": 4.874339580535889, + "learning_rate": 1e-06, + "loss": 0.4163, + "mean_token_accuracy": 0.8339881896972656, + "num_tokens": 5335507.0, + "step": 738 + }, + { + "epoch": 0.7533129459734964, + "grad_norm": 5.8733744621276855, + "learning_rate": 1e-06, + "loss": 0.3968, + "mean_token_accuracy": 0.8314095735549927, + "num_tokens": 5342182.0, + "step": 739 + }, + { + "epoch": 0.7543323139653415, + "grad_norm": 4.769775867462158, + "learning_rate": 1e-06, + "loss": 0.3258, + "mean_token_accuracy": 0.8673291802406311, + "num_tokens": 5351176.0, + "step": 740 + }, + { + "epoch": 0.7553516819571865, + "grad_norm": 5.597282409667969, + "learning_rate": 1e-06, + "loss": 0.4329, + "mean_token_accuracy": 0.8318551182746887, + "num_tokens": 5359169.0, + "step": 741 + }, + { + "epoch": 0.7563710499490316, + "grad_norm": 5.1640944480896, + "learning_rate": 1e-06, + "loss": 0.3957, + "mean_token_accuracy": 0.8534056544303894, + "num_tokens": 5366134.0, + "step": 742 + }, + { + "epoch": 0.7573904179408767, + "grad_norm": 6.200253009796143, + "learning_rate": 1e-06, + "loss": 0.4106, + "mean_token_accuracy": 0.8397390246391296, + "num_tokens": 5371323.0, + "step": 743 + }, + { + "epoch": 0.7584097859327217, + "grad_norm": 5.2570576667785645, + "learning_rate": 1e-06, + "loss": 0.4031, + "mean_token_accuracy": 0.8381760716438293, + "num_tokens": 5378054.0, + "step": 744 + }, + { + "epoch": 0.7594291539245668, + "grad_norm": 5.484820365905762, + "learning_rate": 1e-06, + "loss": 0.4154, + "mean_token_accuracy": 0.8395543098449707, + "num_tokens": 5385298.0, + "step": 745 + }, + { + "epoch": 0.7604485219164119, + "grad_norm": 5.416621208190918, + "learning_rate": 1e-06, + "loss": 0.387, + "mean_token_accuracy": 0.8489913940429688, + "num_tokens": 5391916.0, + "step": 746 + }, + { + "epoch": 0.7614678899082569, + "grad_norm": 4.8919572830200195, + "learning_rate": 1e-06, + "loss": 0.4178, + "mean_token_accuracy": 0.836374819278717, + "num_tokens": 5399656.0, + "step": 747 + }, + { + "epoch": 0.762487257900102, + "grad_norm": 5.390145301818848, + "learning_rate": 1e-06, + "loss": 0.4385, + "mean_token_accuracy": 0.8247775435447693, + "num_tokens": 5406480.0, + "step": 748 + }, + { + "epoch": 0.763506625891947, + "grad_norm": 5.130188941955566, + "learning_rate": 1e-06, + "loss": 0.4111, + "mean_token_accuracy": 0.8455895185470581, + "num_tokens": 5414353.0, + "step": 749 + }, + { + "epoch": 0.764525993883792, + "grad_norm": 6.222805976867676, + "learning_rate": 1e-06, + "loss": 0.4129, + "mean_token_accuracy": 0.8340867161750793, + "num_tokens": 5419937.0, + "step": 750 + }, + { + "epoch": 0.7655453618756372, + "grad_norm": 5.552948474884033, + "learning_rate": 1e-06, + "loss": 0.408, + "mean_token_accuracy": 0.8411045074462891, + "num_tokens": 5426933.0, + "step": 751 + }, + { + "epoch": 0.7665647298674821, + "grad_norm": 4.919328212738037, + "learning_rate": 1e-06, + "loss": 0.359, + "mean_token_accuracy": 0.8507027626037598, + "num_tokens": 5433847.0, + "step": 752 + }, + { + "epoch": 0.7675840978593272, + "grad_norm": 5.573798179626465, + "learning_rate": 1e-06, + "loss": 0.4021, + "mean_token_accuracy": 0.8463443517684937, + "num_tokens": 5440589.0, + "step": 753 + }, + { + "epoch": 0.7686034658511722, + "grad_norm": 5.894457817077637, + "learning_rate": 1e-06, + "loss": 0.4061, + "mean_token_accuracy": 0.8333948850631714, + "num_tokens": 5446787.0, + "step": 754 + }, + { + "epoch": 0.7696228338430173, + "grad_norm": 5.314264297485352, + "learning_rate": 1e-06, + "loss": 0.3989, + "mean_token_accuracy": 0.8342466354370117, + "num_tokens": 5454541.0, + "step": 755 + }, + { + "epoch": 0.7706422018348624, + "grad_norm": 5.290583610534668, + "learning_rate": 1e-06, + "loss": 0.4232, + "mean_token_accuracy": 0.831528902053833, + "num_tokens": 5462129.0, + "step": 756 + }, + { + "epoch": 0.7716615698267074, + "grad_norm": 6.598824501037598, + "learning_rate": 1e-06, + "loss": 0.4343, + "mean_token_accuracy": 0.8336151242256165, + "num_tokens": 5467959.0, + "step": 757 + }, + { + "epoch": 0.7726809378185525, + "grad_norm": 5.2638397216796875, + "learning_rate": 1e-06, + "loss": 0.3582, + "mean_token_accuracy": 0.8483022451400757, + "num_tokens": 5474718.0, + "step": 758 + }, + { + "epoch": 0.7737003058103975, + "grad_norm": 5.764206409454346, + "learning_rate": 1e-06, + "loss": 0.3932, + "mean_token_accuracy": 0.8508241176605225, + "num_tokens": 5481772.0, + "step": 759 + }, + { + "epoch": 0.7747196738022426, + "grad_norm": 6.1513471603393555, + "learning_rate": 1e-06, + "loss": 0.4106, + "mean_token_accuracy": 0.8403687477111816, + "num_tokens": 5488012.0, + "step": 760 + }, + { + "epoch": 0.7757390417940877, + "grad_norm": 5.679276943206787, + "learning_rate": 1e-06, + "loss": 0.4244, + "mean_token_accuracy": 0.8328000903129578, + "num_tokens": 5494920.0, + "step": 761 + }, + { + "epoch": 0.7767584097859327, + "grad_norm": 5.780697822570801, + "learning_rate": 1e-06, + "loss": 0.383, + "mean_token_accuracy": 0.8505288362503052, + "num_tokens": 5501974.0, + "step": 762 + }, + { + "epoch": 0.7777777777777778, + "grad_norm": 6.151556015014648, + "learning_rate": 1e-06, + "loss": 0.4247, + "mean_token_accuracy": 0.8324102163314819, + "num_tokens": 5509000.0, + "step": 763 + }, + { + "epoch": 0.7787971457696228, + "grad_norm": 5.763988018035889, + "learning_rate": 1e-06, + "loss": 0.4073, + "mean_token_accuracy": 0.8417263627052307, + "num_tokens": 5516420.0, + "step": 764 + }, + { + "epoch": 0.7798165137614679, + "grad_norm": 5.575955867767334, + "learning_rate": 1e-06, + "loss": 0.4059, + "mean_token_accuracy": 0.8427504301071167, + "num_tokens": 5524060.0, + "step": 765 + }, + { + "epoch": 0.780835881753313, + "grad_norm": 5.412355422973633, + "learning_rate": 1e-06, + "loss": 0.3848, + "mean_token_accuracy": 0.8435567617416382, + "num_tokens": 5531769.0, + "step": 766 + }, + { + "epoch": 0.781855249745158, + "grad_norm": 5.505460739135742, + "learning_rate": 1e-06, + "loss": 0.3966, + "mean_token_accuracy": 0.847154974937439, + "num_tokens": 5538685.0, + "step": 767 + }, + { + "epoch": 0.7828746177370031, + "grad_norm": 5.0439133644104, + "learning_rate": 1e-06, + "loss": 0.3747, + "mean_token_accuracy": 0.8466494679450989, + "num_tokens": 5546904.0, + "step": 768 + }, + { + "epoch": 0.7838939857288482, + "grad_norm": 5.170846462249756, + "learning_rate": 1e-06, + "loss": 0.3902, + "mean_token_accuracy": 0.8386231660842896, + "num_tokens": 5553975.0, + "step": 769 + }, + { + "epoch": 0.7849133537206932, + "grad_norm": 5.558985710144043, + "learning_rate": 1e-06, + "loss": 0.4332, + "mean_token_accuracy": 0.8346375226974487, + "num_tokens": 5561435.0, + "step": 770 + }, + { + "epoch": 0.7859327217125383, + "grad_norm": 5.077001094818115, + "learning_rate": 1e-06, + "loss": 0.3497, + "mean_token_accuracy": 0.8585691452026367, + "num_tokens": 5568317.0, + "step": 771 + }, + { + "epoch": 0.7869520897043832, + "grad_norm": 5.202419281005859, + "learning_rate": 1e-06, + "loss": 0.3964, + "mean_token_accuracy": 0.8443496227264404, + "num_tokens": 5575242.0, + "step": 772 + }, + { + "epoch": 0.7879714576962283, + "grad_norm": 5.453933238983154, + "learning_rate": 1e-06, + "loss": 0.3894, + "mean_token_accuracy": 0.8467515707015991, + "num_tokens": 5582359.0, + "step": 773 + }, + { + "epoch": 0.7889908256880734, + "grad_norm": 5.259853363037109, + "learning_rate": 1e-06, + "loss": 0.4367, + "mean_token_accuracy": 0.8257154226303101, + "num_tokens": 5589297.0, + "step": 774 + }, + { + "epoch": 0.7900101936799184, + "grad_norm": 5.023660182952881, + "learning_rate": 1e-06, + "loss": 0.3982, + "mean_token_accuracy": 0.8475692272186279, + "num_tokens": 5597212.0, + "step": 775 + }, + { + "epoch": 0.7910295616717635, + "grad_norm": 4.935166835784912, + "learning_rate": 1e-06, + "loss": 0.4, + "mean_token_accuracy": 0.8370544910430908, + "num_tokens": 5605094.0, + "step": 776 + }, + { + "epoch": 0.7920489296636085, + "grad_norm": 4.690821647644043, + "learning_rate": 1e-06, + "loss": 0.3951, + "mean_token_accuracy": 0.8377053141593933, + "num_tokens": 5613245.0, + "step": 777 + }, + { + "epoch": 0.7930682976554536, + "grad_norm": 5.989612579345703, + "learning_rate": 1e-06, + "loss": 0.4351, + "mean_token_accuracy": 0.8359227776527405, + "num_tokens": 5620467.0, + "step": 778 + }, + { + "epoch": 0.7940876656472987, + "grad_norm": 6.621601104736328, + "learning_rate": 1e-06, + "loss": 0.4137, + "mean_token_accuracy": 0.8330037593841553, + "num_tokens": 5627466.0, + "step": 779 + }, + { + "epoch": 0.7951070336391437, + "grad_norm": 5.8139214515686035, + "learning_rate": 1e-06, + "loss": 0.3566, + "mean_token_accuracy": 0.8583642244338989, + "num_tokens": 5634350.0, + "step": 780 + }, + { + "epoch": 0.7961264016309888, + "grad_norm": 4.861350059509277, + "learning_rate": 1e-06, + "loss": 0.3684, + "mean_token_accuracy": 0.8552824854850769, + "num_tokens": 5642029.0, + "step": 781 + }, + { + "epoch": 0.7971457696228338, + "grad_norm": 6.738461971282959, + "learning_rate": 1e-06, + "loss": 0.4495, + "mean_token_accuracy": 0.8210768103599548, + "num_tokens": 5646943.0, + "step": 782 + }, + { + "epoch": 0.7981651376146789, + "grad_norm": 5.018866539001465, + "learning_rate": 1e-06, + "loss": 0.3761, + "mean_token_accuracy": 0.8464555740356445, + "num_tokens": 5654769.0, + "step": 783 + }, + { + "epoch": 0.799184505606524, + "grad_norm": 6.276345729827881, + "learning_rate": 1e-06, + "loss": 0.403, + "mean_token_accuracy": 0.841702938079834, + "num_tokens": 5660476.0, + "step": 784 + }, + { + "epoch": 0.800203873598369, + "grad_norm": 4.5951032638549805, + "learning_rate": 1e-06, + "loss": 0.39, + "mean_token_accuracy": 0.8438402414321899, + "num_tokens": 5670293.0, + "step": 785 + }, + { + "epoch": 0.8012232415902141, + "grad_norm": 4.657723426818848, + "learning_rate": 1e-06, + "loss": 0.3915, + "mean_token_accuracy": 0.8402646780014038, + "num_tokens": 5679039.0, + "step": 786 + }, + { + "epoch": 0.8022426095820592, + "grad_norm": 5.051265239715576, + "learning_rate": 1e-06, + "loss": 0.397, + "mean_token_accuracy": 0.8422866463661194, + "num_tokens": 5686727.0, + "step": 787 + }, + { + "epoch": 0.8032619775739042, + "grad_norm": 6.111390113830566, + "learning_rate": 1e-06, + "loss": 0.3837, + "mean_token_accuracy": 0.8494436740875244, + "num_tokens": 5692790.0, + "step": 788 + }, + { + "epoch": 0.8042813455657493, + "grad_norm": 4.270552158355713, + "learning_rate": 1e-06, + "loss": 0.3864, + "mean_token_accuracy": 0.8476147651672363, + "num_tokens": 5702650.0, + "step": 789 + }, + { + "epoch": 0.8053007135575942, + "grad_norm": 6.334684371948242, + "learning_rate": 1e-06, + "loss": 0.4345, + "mean_token_accuracy": 0.8317856788635254, + "num_tokens": 5708375.0, + "step": 790 + }, + { + "epoch": 0.8063200815494393, + "grad_norm": 4.872244834899902, + "learning_rate": 1e-06, + "loss": 0.4082, + "mean_token_accuracy": 0.840693473815918, + "num_tokens": 5716422.0, + "step": 791 + }, + { + "epoch": 0.8073394495412844, + "grad_norm": 5.041649341583252, + "learning_rate": 1e-06, + "loss": 0.4129, + "mean_token_accuracy": 0.8393846154212952, + "num_tokens": 5724675.0, + "step": 792 + }, + { + "epoch": 0.8083588175331294, + "grad_norm": 5.351203918457031, + "learning_rate": 1e-06, + "loss": 0.4004, + "mean_token_accuracy": 0.8439068794250488, + "num_tokens": 5732262.0, + "step": 793 + }, + { + "epoch": 0.8093781855249745, + "grad_norm": 4.809719562530518, + "learning_rate": 1e-06, + "loss": 0.4041, + "mean_token_accuracy": 0.8369211554527283, + "num_tokens": 5740152.0, + "step": 794 + }, + { + "epoch": 0.8103975535168195, + "grad_norm": 5.701815128326416, + "learning_rate": 1e-06, + "loss": 0.4027, + "mean_token_accuracy": 0.8440060615539551, + "num_tokens": 5746976.0, + "step": 795 + }, + { + "epoch": 0.8114169215086646, + "grad_norm": 5.323419094085693, + "learning_rate": 1e-06, + "loss": 0.4275, + "mean_token_accuracy": 0.8254895806312561, + "num_tokens": 5753838.0, + "step": 796 + }, + { + "epoch": 0.8124362895005097, + "grad_norm": 5.079898834228516, + "learning_rate": 1e-06, + "loss": 0.4182, + "mean_token_accuracy": 0.8343514800071716, + "num_tokens": 5762012.0, + "step": 797 + }, + { + "epoch": 0.8134556574923547, + "grad_norm": 5.510432243347168, + "learning_rate": 1e-06, + "loss": 0.4143, + "mean_token_accuracy": 0.8376230001449585, + "num_tokens": 5769475.0, + "step": 798 + }, + { + "epoch": 0.8144750254841998, + "grad_norm": 5.1207146644592285, + "learning_rate": 1e-06, + "loss": 0.4032, + "mean_token_accuracy": 0.8481038808822632, + "num_tokens": 5776914.0, + "step": 799 + }, + { + "epoch": 0.8154943934760448, + "grad_norm": 5.357790946960449, + "learning_rate": 1e-06, + "loss": 0.4469, + "mean_token_accuracy": 0.8266454935073853, + "num_tokens": 5785261.0, + "step": 800 + }, + { + "epoch": 0.8165137614678899, + "grad_norm": 6.0335373878479, + "learning_rate": 1e-06, + "loss": 0.3949, + "mean_token_accuracy": 0.8412995338439941, + "num_tokens": 5791101.0, + "step": 801 + }, + { + "epoch": 0.817533129459735, + "grad_norm": 6.00661563873291, + "learning_rate": 1e-06, + "loss": 0.4139, + "mean_token_accuracy": 0.8355122804641724, + "num_tokens": 5797248.0, + "step": 802 + }, + { + "epoch": 0.81855249745158, + "grad_norm": 5.090859413146973, + "learning_rate": 1e-06, + "loss": 0.3832, + "mean_token_accuracy": 0.8536157608032227, + "num_tokens": 5804805.0, + "step": 803 + }, + { + "epoch": 0.8195718654434251, + "grad_norm": 5.373654365539551, + "learning_rate": 1e-06, + "loss": 0.3983, + "mean_token_accuracy": 0.8446439504623413, + "num_tokens": 5812279.0, + "step": 804 + }, + { + "epoch": 0.8205912334352702, + "grad_norm": 4.813554286956787, + "learning_rate": 1e-06, + "loss": 0.3784, + "mean_token_accuracy": 0.8554280400276184, + "num_tokens": 5820609.0, + "step": 805 + }, + { + "epoch": 0.8216106014271152, + "grad_norm": 4.5580878257751465, + "learning_rate": 1e-06, + "loss": 0.3832, + "mean_token_accuracy": 0.8488964438438416, + "num_tokens": 5829773.0, + "step": 806 + }, + { + "epoch": 0.8226299694189603, + "grad_norm": 6.083618640899658, + "learning_rate": 1e-06, + "loss": 0.4611, + "mean_token_accuracy": 0.8274751901626587, + "num_tokens": 5836012.0, + "step": 807 + }, + { + "epoch": 0.8236493374108053, + "grad_norm": 5.565022945404053, + "learning_rate": 1e-06, + "loss": 0.3966, + "mean_token_accuracy": 0.8349344730377197, + "num_tokens": 5842988.0, + "step": 808 + }, + { + "epoch": 0.8246687054026504, + "grad_norm": 5.381558418273926, + "learning_rate": 1e-06, + "loss": 0.4286, + "mean_token_accuracy": 0.8342804908752441, + "num_tokens": 5849415.0, + "step": 809 + }, + { + "epoch": 0.8256880733944955, + "grad_norm": 6.025678634643555, + "learning_rate": 1e-06, + "loss": 0.4059, + "mean_token_accuracy": 0.8369750380516052, + "num_tokens": 5855723.0, + "step": 810 + }, + { + "epoch": 0.8267074413863404, + "grad_norm": 5.989630222320557, + "learning_rate": 1e-06, + "loss": 0.4168, + "mean_token_accuracy": 0.8419371843338013, + "num_tokens": 5861829.0, + "step": 811 + }, + { + "epoch": 0.8277268093781855, + "grad_norm": 5.548488140106201, + "learning_rate": 1e-06, + "loss": 0.4038, + "mean_token_accuracy": 0.8402601480484009, + "num_tokens": 5869285.0, + "step": 812 + }, + { + "epoch": 0.8287461773700305, + "grad_norm": 4.656595706939697, + "learning_rate": 1e-06, + "loss": 0.4037, + "mean_token_accuracy": 0.8403873443603516, + "num_tokens": 5877621.0, + "step": 813 + }, + { + "epoch": 0.8297655453618756, + "grad_norm": 5.540914535522461, + "learning_rate": 1e-06, + "loss": 0.3786, + "mean_token_accuracy": 0.8514501452445984, + "num_tokens": 5885280.0, + "step": 814 + }, + { + "epoch": 0.8307849133537207, + "grad_norm": 6.2540364265441895, + "learning_rate": 1e-06, + "loss": 0.4205, + "mean_token_accuracy": 0.8306045532226562, + "num_tokens": 5892015.0, + "step": 815 + }, + { + "epoch": 0.8318042813455657, + "grad_norm": 4.31207799911499, + "learning_rate": 1e-06, + "loss": 0.3362, + "mean_token_accuracy": 0.8525002598762512, + "num_tokens": 5901505.0, + "step": 816 + }, + { + "epoch": 0.8328236493374108, + "grad_norm": 5.16297721862793, + "learning_rate": 1e-06, + "loss": 0.3759, + "mean_token_accuracy": 0.8492979407310486, + "num_tokens": 5909071.0, + "step": 817 + }, + { + "epoch": 0.8338430173292558, + "grad_norm": 5.731659412384033, + "learning_rate": 1e-06, + "loss": 0.4247, + "mean_token_accuracy": 0.8306224346160889, + "num_tokens": 5916178.0, + "step": 818 + }, + { + "epoch": 0.8348623853211009, + "grad_norm": 6.059806823730469, + "learning_rate": 1e-06, + "loss": 0.4147, + "mean_token_accuracy": 0.8310813903808594, + "num_tokens": 5922438.0, + "step": 819 + }, + { + "epoch": 0.835881753312946, + "grad_norm": 5.986852169036865, + "learning_rate": 1e-06, + "loss": 0.4025, + "mean_token_accuracy": 0.8338366746902466, + "num_tokens": 5928529.0, + "step": 820 + }, + { + "epoch": 0.836901121304791, + "grad_norm": 5.535304546356201, + "learning_rate": 1e-06, + "loss": 0.3966, + "mean_token_accuracy": 0.8361024260520935, + "num_tokens": 5934946.0, + "step": 821 + }, + { + "epoch": 0.8379204892966361, + "grad_norm": 6.147219657897949, + "learning_rate": 1e-06, + "loss": 0.4214, + "mean_token_accuracy": 0.836172342300415, + "num_tokens": 5940510.0, + "step": 822 + }, + { + "epoch": 0.8389398572884812, + "grad_norm": 4.516496658325195, + "learning_rate": 1e-06, + "loss": 0.3316, + "mean_token_accuracy": 0.8592969179153442, + "num_tokens": 5948610.0, + "step": 823 + }, + { + "epoch": 0.8399592252803262, + "grad_norm": 5.43679141998291, + "learning_rate": 1e-06, + "loss": 0.3833, + "mean_token_accuracy": 0.843454122543335, + "num_tokens": 5956210.0, + "step": 824 + }, + { + "epoch": 0.8409785932721713, + "grad_norm": 5.771844387054443, + "learning_rate": 1e-06, + "loss": 0.4329, + "mean_token_accuracy": 0.828596830368042, + "num_tokens": 5962825.0, + "step": 825 + }, + { + "epoch": 0.8419979612640163, + "grad_norm": 5.290629863739014, + "learning_rate": 1e-06, + "loss": 0.4406, + "mean_token_accuracy": 0.8269610404968262, + "num_tokens": 5970868.0, + "step": 826 + }, + { + "epoch": 0.8430173292558614, + "grad_norm": 5.445380687713623, + "learning_rate": 1e-06, + "loss": 0.4062, + "mean_token_accuracy": 0.8387771844863892, + "num_tokens": 5978427.0, + "step": 827 + }, + { + "epoch": 0.8440366972477065, + "grad_norm": 4.9413909912109375, + "learning_rate": 1e-06, + "loss": 0.4219, + "mean_token_accuracy": 0.8381062150001526, + "num_tokens": 5986280.0, + "step": 828 + }, + { + "epoch": 0.8450560652395515, + "grad_norm": 4.815649509429932, + "learning_rate": 1e-06, + "loss": 0.3619, + "mean_token_accuracy": 0.8540675044059753, + "num_tokens": 5993811.0, + "step": 829 + }, + { + "epoch": 0.8460754332313966, + "grad_norm": 5.104669570922852, + "learning_rate": 1e-06, + "loss": 0.4048, + "mean_token_accuracy": 0.8445725440979004, + "num_tokens": 6001425.0, + "step": 830 + }, + { + "epoch": 0.8470948012232415, + "grad_norm": 4.271953582763672, + "learning_rate": 1e-06, + "loss": 0.3481, + "mean_token_accuracy": 0.8576292991638184, + "num_tokens": 6010262.0, + "step": 831 + }, + { + "epoch": 0.8481141692150866, + "grad_norm": 5.2875471115112305, + "learning_rate": 1e-06, + "loss": 0.4274, + "mean_token_accuracy": 0.8316732048988342, + "num_tokens": 6018392.0, + "step": 832 + }, + { + "epoch": 0.8491335372069317, + "grad_norm": 6.319366455078125, + "learning_rate": 1e-06, + "loss": 0.4173, + "mean_token_accuracy": 0.8299537301063538, + "num_tokens": 6023869.0, + "step": 833 + }, + { + "epoch": 0.8501529051987767, + "grad_norm": 5.139124393463135, + "learning_rate": 1e-06, + "loss": 0.3837, + "mean_token_accuracy": 0.8470261693000793, + "num_tokens": 6031437.0, + "step": 834 + }, + { + "epoch": 0.8511722731906218, + "grad_norm": 5.539470195770264, + "learning_rate": 1e-06, + "loss": 0.4239, + "mean_token_accuracy": 0.8358088731765747, + "num_tokens": 6037733.0, + "step": 835 + }, + { + "epoch": 0.8521916411824668, + "grad_norm": 5.764766693115234, + "learning_rate": 1e-06, + "loss": 0.414, + "mean_token_accuracy": 0.8388986587524414, + "num_tokens": 6043894.0, + "step": 836 + }, + { + "epoch": 0.8532110091743119, + "grad_norm": 4.951055526733398, + "learning_rate": 1e-06, + "loss": 0.3503, + "mean_token_accuracy": 0.8618273138999939, + "num_tokens": 6051897.0, + "step": 837 + }, + { + "epoch": 0.854230377166157, + "grad_norm": 5.472519874572754, + "learning_rate": 1e-06, + "loss": 0.4051, + "mean_token_accuracy": 0.8333524465560913, + "num_tokens": 6058180.0, + "step": 838 + }, + { + "epoch": 0.855249745158002, + "grad_norm": 5.416780948638916, + "learning_rate": 1e-06, + "loss": 0.3674, + "mean_token_accuracy": 0.854871392250061, + "num_tokens": 6065828.0, + "step": 839 + }, + { + "epoch": 0.8562691131498471, + "grad_norm": 5.955313682556152, + "learning_rate": 1e-06, + "loss": 0.4094, + "mean_token_accuracy": 0.8395602703094482, + "num_tokens": 6073061.0, + "step": 840 + }, + { + "epoch": 0.8572884811416922, + "grad_norm": 4.912600040435791, + "learning_rate": 1e-06, + "loss": 0.3796, + "mean_token_accuracy": 0.8489820957183838, + "num_tokens": 6081357.0, + "step": 841 + }, + { + "epoch": 0.8583078491335372, + "grad_norm": 5.486711025238037, + "learning_rate": 1e-06, + "loss": 0.4187, + "mean_token_accuracy": 0.8316606283187866, + "num_tokens": 6088042.0, + "step": 842 + }, + { + "epoch": 0.8593272171253823, + "grad_norm": 5.369353294372559, + "learning_rate": 1e-06, + "loss": 0.3921, + "mean_token_accuracy": 0.8448452353477478, + "num_tokens": 6094982.0, + "step": 843 + }, + { + "epoch": 0.8603465851172273, + "grad_norm": 5.373276233673096, + "learning_rate": 1e-06, + "loss": 0.364, + "mean_token_accuracy": 0.8536550998687744, + "num_tokens": 6101691.0, + "step": 844 + }, + { + "epoch": 0.8613659531090724, + "grad_norm": 6.11181640625, + "learning_rate": 1e-06, + "loss": 0.4192, + "mean_token_accuracy": 0.8330100774765015, + "num_tokens": 6108026.0, + "step": 845 + }, + { + "epoch": 0.8623853211009175, + "grad_norm": 5.855648040771484, + "learning_rate": 1e-06, + "loss": 0.4103, + "mean_token_accuracy": 0.8416413068771362, + "num_tokens": 6114503.0, + "step": 846 + }, + { + "epoch": 0.8634046890927625, + "grad_norm": 4.88825798034668, + "learning_rate": 1e-06, + "loss": 0.4217, + "mean_token_accuracy": 0.8353726863861084, + "num_tokens": 6121929.0, + "step": 847 + }, + { + "epoch": 0.8644240570846076, + "grad_norm": 5.677170753479004, + "learning_rate": 1e-06, + "loss": 0.3746, + "mean_token_accuracy": 0.8485962748527527, + "num_tokens": 6128435.0, + "step": 848 + }, + { + "epoch": 0.8654434250764526, + "grad_norm": 4.367680072784424, + "learning_rate": 1e-06, + "loss": 0.371, + "mean_token_accuracy": 0.8529970645904541, + "num_tokens": 6136834.0, + "step": 849 + }, + { + "epoch": 0.8664627930682977, + "grad_norm": 5.174642562866211, + "learning_rate": 1e-06, + "loss": 0.409, + "mean_token_accuracy": 0.8363621830940247, + "num_tokens": 6144700.0, + "step": 850 + }, + { + "epoch": 0.8674821610601428, + "grad_norm": 4.574234962463379, + "learning_rate": 1e-06, + "loss": 0.3402, + "mean_token_accuracy": 0.8621616959571838, + "num_tokens": 6153635.0, + "step": 851 + }, + { + "epoch": 0.8685015290519877, + "grad_norm": 4.910515308380127, + "learning_rate": 1e-06, + "loss": 0.3849, + "mean_token_accuracy": 0.8451647758483887, + "num_tokens": 6161737.0, + "step": 852 + }, + { + "epoch": 0.8695208970438328, + "grad_norm": 4.977810859680176, + "learning_rate": 1e-06, + "loss": 0.3769, + "mean_token_accuracy": 0.8496698141098022, + "num_tokens": 6169160.0, + "step": 853 + }, + { + "epoch": 0.8705402650356778, + "grad_norm": 5.007872104644775, + "learning_rate": 1e-06, + "loss": 0.3458, + "mean_token_accuracy": 0.8586136102676392, + "num_tokens": 6176161.0, + "step": 854 + }, + { + "epoch": 0.8715596330275229, + "grad_norm": 4.941714763641357, + "learning_rate": 1e-06, + "loss": 0.3983, + "mean_token_accuracy": 0.8354116082191467, + "num_tokens": 6184311.0, + "step": 855 + }, + { + "epoch": 0.872579001019368, + "grad_norm": 5.341039180755615, + "learning_rate": 1e-06, + "loss": 0.4025, + "mean_token_accuracy": 0.841097891330719, + "num_tokens": 6191964.0, + "step": 856 + }, + { + "epoch": 0.873598369011213, + "grad_norm": 4.743738651275635, + "learning_rate": 1e-06, + "loss": 0.3827, + "mean_token_accuracy": 0.8487321734428406, + "num_tokens": 6199735.0, + "step": 857 + }, + { + "epoch": 0.8746177370030581, + "grad_norm": 5.506189823150635, + "learning_rate": 1e-06, + "loss": 0.4102, + "mean_token_accuracy": 0.8368375301361084, + "num_tokens": 6207100.0, + "step": 858 + }, + { + "epoch": 0.8756371049949032, + "grad_norm": 6.073219299316406, + "learning_rate": 1e-06, + "loss": 0.4467, + "mean_token_accuracy": 0.8225088119506836, + "num_tokens": 6212902.0, + "step": 859 + }, + { + "epoch": 0.8766564729867482, + "grad_norm": 5.082705974578857, + "learning_rate": 1e-06, + "loss": 0.3894, + "mean_token_accuracy": 0.8472175598144531, + "num_tokens": 6220037.0, + "step": 860 + }, + { + "epoch": 0.8776758409785933, + "grad_norm": 6.20258092880249, + "learning_rate": 1e-06, + "loss": 0.4139, + "mean_token_accuracy": 0.8355940580368042, + "num_tokens": 6226301.0, + "step": 861 + }, + { + "epoch": 0.8786952089704383, + "grad_norm": 4.923445701599121, + "learning_rate": 1e-06, + "loss": 0.3897, + "mean_token_accuracy": 0.8416748046875, + "num_tokens": 6234612.0, + "step": 862 + }, + { + "epoch": 0.8797145769622834, + "grad_norm": 5.823384761810303, + "learning_rate": 1e-06, + "loss": 0.3856, + "mean_token_accuracy": 0.8488984107971191, + "num_tokens": 6241340.0, + "step": 863 + }, + { + "epoch": 0.8807339449541285, + "grad_norm": 5.7776923179626465, + "learning_rate": 1e-06, + "loss": 0.4609, + "mean_token_accuracy": 0.8214945793151855, + "num_tokens": 6248025.0, + "step": 864 + }, + { + "epoch": 0.8817533129459735, + "grad_norm": 5.503491401672363, + "learning_rate": 1e-06, + "loss": 0.4176, + "mean_token_accuracy": 0.8301821947097778, + "num_tokens": 6254338.0, + "step": 865 + }, + { + "epoch": 0.8827726809378186, + "grad_norm": 4.896544933319092, + "learning_rate": 1e-06, + "loss": 0.3868, + "mean_token_accuracy": 0.8538063764572144, + "num_tokens": 6261875.0, + "step": 866 + }, + { + "epoch": 0.8837920489296636, + "grad_norm": 5.2178955078125, + "learning_rate": 1e-06, + "loss": 0.4064, + "mean_token_accuracy": 0.8431545495986938, + "num_tokens": 6269717.0, + "step": 867 + }, + { + "epoch": 0.8848114169215087, + "grad_norm": 4.889104843139648, + "learning_rate": 1e-06, + "loss": 0.4043, + "mean_token_accuracy": 0.8393422365188599, + "num_tokens": 6277660.0, + "step": 868 + }, + { + "epoch": 0.8858307849133538, + "grad_norm": 5.540349006652832, + "learning_rate": 1e-06, + "loss": 0.3671, + "mean_token_accuracy": 0.8529056310653687, + "num_tokens": 6284502.0, + "step": 869 + }, + { + "epoch": 0.8868501529051988, + "grad_norm": 5.319561958312988, + "learning_rate": 1e-06, + "loss": 0.4076, + "mean_token_accuracy": 0.8401526212692261, + "num_tokens": 6291916.0, + "step": 870 + }, + { + "epoch": 0.8878695208970439, + "grad_norm": 4.923853874206543, + "learning_rate": 1e-06, + "loss": 0.3981, + "mean_token_accuracy": 0.8442630767822266, + "num_tokens": 6299622.0, + "step": 871 + }, + { + "epoch": 0.8888888888888888, + "grad_norm": 5.842583179473877, + "learning_rate": 1e-06, + "loss": 0.3926, + "mean_token_accuracy": 0.8412972092628479, + "num_tokens": 6305758.0, + "step": 872 + }, + { + "epoch": 0.8899082568807339, + "grad_norm": 5.1663103103637695, + "learning_rate": 1e-06, + "loss": 0.4083, + "mean_token_accuracy": 0.8404912948608398, + "num_tokens": 6315017.0, + "step": 873 + }, + { + "epoch": 0.890927624872579, + "grad_norm": 5.73947286605835, + "learning_rate": 1e-06, + "loss": 0.4212, + "mean_token_accuracy": 0.8309199810028076, + "num_tokens": 6321538.0, + "step": 874 + }, + { + "epoch": 0.891946992864424, + "grad_norm": 5.037108898162842, + "learning_rate": 1e-06, + "loss": 0.3789, + "mean_token_accuracy": 0.8488031029701233, + "num_tokens": 6329418.0, + "step": 875 + }, + { + "epoch": 0.8929663608562691, + "grad_norm": 6.172038555145264, + "learning_rate": 1e-06, + "loss": 0.4489, + "mean_token_accuracy": 0.8234406113624573, + "num_tokens": 6336448.0, + "step": 876 + }, + { + "epoch": 0.8939857288481141, + "grad_norm": 5.270371437072754, + "learning_rate": 1e-06, + "loss": 0.4237, + "mean_token_accuracy": 0.8369457125663757, + "num_tokens": 6344900.0, + "step": 877 + }, + { + "epoch": 0.8950050968399592, + "grad_norm": 5.138740539550781, + "learning_rate": 1e-06, + "loss": 0.3796, + "mean_token_accuracy": 0.8489075899124146, + "num_tokens": 6352218.0, + "step": 878 + }, + { + "epoch": 0.8960244648318043, + "grad_norm": 5.54639196395874, + "learning_rate": 1e-06, + "loss": 0.4073, + "mean_token_accuracy": 0.8374179601669312, + "num_tokens": 6358702.0, + "step": 879 + }, + { + "epoch": 0.8970438328236493, + "grad_norm": 4.935513019561768, + "learning_rate": 1e-06, + "loss": 0.3723, + "mean_token_accuracy": 0.855151355266571, + "num_tokens": 6366858.0, + "step": 880 + }, + { + "epoch": 0.8980632008154944, + "grad_norm": 5.399344444274902, + "learning_rate": 1e-06, + "loss": 0.3993, + "mean_token_accuracy": 0.8425099849700928, + "num_tokens": 6373316.0, + "step": 881 + }, + { + "epoch": 0.8990825688073395, + "grad_norm": 5.3485822677612305, + "learning_rate": 1e-06, + "loss": 0.4031, + "mean_token_accuracy": 0.8442529439926147, + "num_tokens": 6380598.0, + "step": 882 + }, + { + "epoch": 0.9001019367991845, + "grad_norm": 6.192961692810059, + "learning_rate": 1e-06, + "loss": 0.4344, + "mean_token_accuracy": 0.8295199275016785, + "num_tokens": 6386219.0, + "step": 883 + }, + { + "epoch": 0.9011213047910296, + "grad_norm": 4.752520561218262, + "learning_rate": 1e-06, + "loss": 0.4427, + "mean_token_accuracy": 0.8277541399002075, + "num_tokens": 6394372.0, + "step": 884 + }, + { + "epoch": 0.9021406727828746, + "grad_norm": 4.927333831787109, + "learning_rate": 1e-06, + "loss": 0.3874, + "mean_token_accuracy": 0.8458511233329773, + "num_tokens": 6402736.0, + "step": 885 + }, + { + "epoch": 0.9031600407747197, + "grad_norm": 4.938380241394043, + "learning_rate": 1e-06, + "loss": 0.3861, + "mean_token_accuracy": 0.8474178314208984, + "num_tokens": 6410483.0, + "step": 886 + }, + { + "epoch": 0.9041794087665648, + "grad_norm": 4.893454551696777, + "learning_rate": 1e-06, + "loss": 0.3644, + "mean_token_accuracy": 0.8584307432174683, + "num_tokens": 6417564.0, + "step": 887 + }, + { + "epoch": 0.9051987767584098, + "grad_norm": 4.804189682006836, + "learning_rate": 1e-06, + "loss": 0.3869, + "mean_token_accuracy": 0.8427751064300537, + "num_tokens": 6425193.0, + "step": 888 + }, + { + "epoch": 0.9062181447502549, + "grad_norm": 5.102541923522949, + "learning_rate": 1e-06, + "loss": 0.3923, + "mean_token_accuracy": 0.8412259221076965, + "num_tokens": 6431959.0, + "step": 889 + }, + { + "epoch": 0.9072375127420998, + "grad_norm": 5.575172424316406, + "learning_rate": 1e-06, + "loss": 0.4282, + "mean_token_accuracy": 0.8341455459594727, + "num_tokens": 6438465.0, + "step": 890 + }, + { + "epoch": 0.908256880733945, + "grad_norm": 6.318997383117676, + "learning_rate": 1e-06, + "loss": 0.4419, + "mean_token_accuracy": 0.8279570937156677, + "num_tokens": 6444412.0, + "step": 891 + }, + { + "epoch": 0.90927624872579, + "grad_norm": 5.118502616882324, + "learning_rate": 1e-06, + "loss": 0.3583, + "mean_token_accuracy": 0.858089804649353, + "num_tokens": 6452294.0, + "step": 892 + }, + { + "epoch": 0.910295616717635, + "grad_norm": 5.31350564956665, + "learning_rate": 1e-06, + "loss": 0.368, + "mean_token_accuracy": 0.8416301012039185, + "num_tokens": 6459149.0, + "step": 893 + }, + { + "epoch": 0.9113149847094801, + "grad_norm": 4.958829402923584, + "learning_rate": 1e-06, + "loss": 0.3787, + "mean_token_accuracy": 0.8542525768280029, + "num_tokens": 6466764.0, + "step": 894 + }, + { + "epoch": 0.9123343527013251, + "grad_norm": 5.223799228668213, + "learning_rate": 1e-06, + "loss": 0.3782, + "mean_token_accuracy": 0.8522741198539734, + "num_tokens": 6473709.0, + "step": 895 + }, + { + "epoch": 0.9133537206931702, + "grad_norm": 5.18353796005249, + "learning_rate": 1e-06, + "loss": 0.3991, + "mean_token_accuracy": 0.8466893434524536, + "num_tokens": 6481666.0, + "step": 896 + }, + { + "epoch": 0.9143730886850153, + "grad_norm": 6.218839168548584, + "learning_rate": 1e-06, + "loss": 0.4243, + "mean_token_accuracy": 0.8318468928337097, + "num_tokens": 6487749.0, + "step": 897 + }, + { + "epoch": 0.9153924566768603, + "grad_norm": 3.9982333183288574, + "learning_rate": 1e-06, + "loss": 0.3091, + "mean_token_accuracy": 0.8611539602279663, + "num_tokens": 6497802.0, + "step": 898 + }, + { + "epoch": 0.9164118246687054, + "grad_norm": 5.8713274002075195, + "learning_rate": 1e-06, + "loss": 0.4169, + "mean_token_accuracy": 0.8362828493118286, + "num_tokens": 6505215.0, + "step": 899 + }, + { + "epoch": 0.9174311926605505, + "grad_norm": 5.671985626220703, + "learning_rate": 1e-06, + "loss": 0.3854, + "mean_token_accuracy": 0.8494772911071777, + "num_tokens": 6512237.0, + "step": 900 + }, + { + "epoch": 0.9184505606523955, + "grad_norm": 5.2495527267456055, + "learning_rate": 1e-06, + "loss": 0.3878, + "mean_token_accuracy": 0.8441956043243408, + "num_tokens": 6519184.0, + "step": 901 + }, + { + "epoch": 0.9194699286442406, + "grad_norm": 5.40781307220459, + "learning_rate": 1e-06, + "loss": 0.4116, + "mean_token_accuracy": 0.8366631865501404, + "num_tokens": 6526198.0, + "step": 902 + }, + { + "epoch": 0.9204892966360856, + "grad_norm": 6.295017719268799, + "learning_rate": 1e-06, + "loss": 0.4585, + "mean_token_accuracy": 0.8304193615913391, + "num_tokens": 6532614.0, + "step": 903 + }, + { + "epoch": 0.9215086646279307, + "grad_norm": 5.944131851196289, + "learning_rate": 1e-06, + "loss": 0.4041, + "mean_token_accuracy": 0.8320270776748657, + "num_tokens": 6539475.0, + "step": 904 + }, + { + "epoch": 0.9225280326197758, + "grad_norm": 5.632084369659424, + "learning_rate": 1e-06, + "loss": 0.4176, + "mean_token_accuracy": 0.838144838809967, + "num_tokens": 6547046.0, + "step": 905 + }, + { + "epoch": 0.9235474006116208, + "grad_norm": 5.947238445281982, + "learning_rate": 1e-06, + "loss": 0.3831, + "mean_token_accuracy": 0.8449617624282837, + "num_tokens": 6554274.0, + "step": 906 + }, + { + "epoch": 0.9245667686034659, + "grad_norm": 5.222867012023926, + "learning_rate": 1e-06, + "loss": 0.3926, + "mean_token_accuracy": 0.8458458185195923, + "num_tokens": 6561196.0, + "step": 907 + }, + { + "epoch": 0.9255861365953109, + "grad_norm": 5.607975006103516, + "learning_rate": 1e-06, + "loss": 0.4145, + "mean_token_accuracy": 0.8416345119476318, + "num_tokens": 6568150.0, + "step": 908 + }, + { + "epoch": 0.926605504587156, + "grad_norm": 4.584492206573486, + "learning_rate": 1e-06, + "loss": 0.3895, + "mean_token_accuracy": 0.8484870195388794, + "num_tokens": 6577245.0, + "step": 909 + }, + { + "epoch": 0.9276248725790011, + "grad_norm": 4.695788860321045, + "learning_rate": 1e-06, + "loss": 0.379, + "mean_token_accuracy": 0.850920557975769, + "num_tokens": 6585930.0, + "step": 910 + }, + { + "epoch": 0.928644240570846, + "grad_norm": 5.26473331451416, + "learning_rate": 1e-06, + "loss": 0.4126, + "mean_token_accuracy": 0.8380853533744812, + "num_tokens": 6592785.0, + "step": 911 + }, + { + "epoch": 0.9296636085626911, + "grad_norm": 5.204003810882568, + "learning_rate": 1e-06, + "loss": 0.4167, + "mean_token_accuracy": 0.8404630422592163, + "num_tokens": 6600132.0, + "step": 912 + }, + { + "epoch": 0.9306829765545361, + "grad_norm": 5.134033203125, + "learning_rate": 1e-06, + "loss": 0.3688, + "mean_token_accuracy": 0.8516488075256348, + "num_tokens": 6606863.0, + "step": 913 + }, + { + "epoch": 0.9317023445463812, + "grad_norm": 4.795969009399414, + "learning_rate": 1e-06, + "loss": 0.3876, + "mean_token_accuracy": 0.8502967357635498, + "num_tokens": 6614695.0, + "step": 914 + }, + { + "epoch": 0.9327217125382263, + "grad_norm": 4.938913822174072, + "learning_rate": 1e-06, + "loss": 0.426, + "mean_token_accuracy": 0.8352881073951721, + "num_tokens": 6624057.0, + "step": 915 + }, + { + "epoch": 0.9337410805300713, + "grad_norm": 4.827045917510986, + "learning_rate": 1e-06, + "loss": 0.4286, + "mean_token_accuracy": 0.8295688629150391, + "num_tokens": 6631896.0, + "step": 916 + }, + { + "epoch": 0.9347604485219164, + "grad_norm": 4.869732856750488, + "learning_rate": 1e-06, + "loss": 0.3793, + "mean_token_accuracy": 0.8530046343803406, + "num_tokens": 6639918.0, + "step": 917 + }, + { + "epoch": 0.9357798165137615, + "grad_norm": 4.069597244262695, + "learning_rate": 1e-06, + "loss": 0.3467, + "mean_token_accuracy": 0.853972852230072, + "num_tokens": 6649621.0, + "step": 918 + }, + { + "epoch": 0.9367991845056065, + "grad_norm": 5.314488887786865, + "learning_rate": 1e-06, + "loss": 0.379, + "mean_token_accuracy": 0.8421745300292969, + "num_tokens": 6656653.0, + "step": 919 + }, + { + "epoch": 0.9378185524974516, + "grad_norm": 5.8823771476745605, + "learning_rate": 1e-06, + "loss": 0.4127, + "mean_token_accuracy": 0.8388401865959167, + "num_tokens": 6662511.0, + "step": 920 + }, + { + "epoch": 0.9388379204892966, + "grad_norm": 5.340895175933838, + "learning_rate": 1e-06, + "loss": 0.381, + "mean_token_accuracy": 0.8491483926773071, + "num_tokens": 6668740.0, + "step": 921 + }, + { + "epoch": 0.9398572884811417, + "grad_norm": 5.009697437286377, + "learning_rate": 1e-06, + "loss": 0.3947, + "mean_token_accuracy": 0.8470079898834229, + "num_tokens": 6675889.0, + "step": 922 + }, + { + "epoch": 0.9408766564729868, + "grad_norm": 5.3810505867004395, + "learning_rate": 1e-06, + "loss": 0.3891, + "mean_token_accuracy": 0.8445613384246826, + "num_tokens": 6682710.0, + "step": 923 + }, + { + "epoch": 0.9418960244648318, + "grad_norm": 5.798577308654785, + "learning_rate": 1e-06, + "loss": 0.4248, + "mean_token_accuracy": 0.8336218595504761, + "num_tokens": 6689047.0, + "step": 924 + }, + { + "epoch": 0.9429153924566769, + "grad_norm": 5.197432994842529, + "learning_rate": 1e-06, + "loss": 0.3762, + "mean_token_accuracy": 0.8474849462509155, + "num_tokens": 6696900.0, + "step": 925 + }, + { + "epoch": 0.9439347604485219, + "grad_norm": 5.470854759216309, + "learning_rate": 1e-06, + "loss": 0.3987, + "mean_token_accuracy": 0.8422573804855347, + "num_tokens": 6703552.0, + "step": 926 + }, + { + "epoch": 0.944954128440367, + "grad_norm": 5.114354133605957, + "learning_rate": 1e-06, + "loss": 0.4004, + "mean_token_accuracy": 0.849012017250061, + "num_tokens": 6710550.0, + "step": 927 + }, + { + "epoch": 0.9459734964322121, + "grad_norm": 6.2023773193359375, + "learning_rate": 1e-06, + "loss": 0.4077, + "mean_token_accuracy": 0.8381273746490479, + "num_tokens": 6716770.0, + "step": 928 + }, + { + "epoch": 0.9469928644240571, + "grad_norm": 5.18090295791626, + "learning_rate": 1e-06, + "loss": 0.4122, + "mean_token_accuracy": 0.8372250199317932, + "num_tokens": 6724501.0, + "step": 929 + }, + { + "epoch": 0.9480122324159022, + "grad_norm": 4.609687328338623, + "learning_rate": 1e-06, + "loss": 0.37, + "mean_token_accuracy": 0.8539831042289734, + "num_tokens": 6733764.0, + "step": 930 + }, + { + "epoch": 0.9490316004077471, + "grad_norm": 4.601118564605713, + "learning_rate": 1e-06, + "loss": 0.3458, + "mean_token_accuracy": 0.860045850276947, + "num_tokens": 6741678.0, + "step": 931 + }, + { + "epoch": 0.9500509683995922, + "grad_norm": 4.906056880950928, + "learning_rate": 1e-06, + "loss": 0.3781, + "mean_token_accuracy": 0.85215163230896, + "num_tokens": 6749116.0, + "step": 932 + }, + { + "epoch": 0.9510703363914373, + "grad_norm": 5.48580265045166, + "learning_rate": 1e-06, + "loss": 0.4116, + "mean_token_accuracy": 0.8362563848495483, + "num_tokens": 6756149.0, + "step": 933 + }, + { + "epoch": 0.9520897043832823, + "grad_norm": 5.1335883140563965, + "learning_rate": 1e-06, + "loss": 0.3724, + "mean_token_accuracy": 0.8510727882385254, + "num_tokens": 6763392.0, + "step": 934 + }, + { + "epoch": 0.9531090723751274, + "grad_norm": 5.072076320648193, + "learning_rate": 1e-06, + "loss": 0.4001, + "mean_token_accuracy": 0.8469406366348267, + "num_tokens": 6771465.0, + "step": 935 + }, + { + "epoch": 0.9541284403669725, + "grad_norm": 5.335959434509277, + "learning_rate": 1e-06, + "loss": 0.3886, + "mean_token_accuracy": 0.8426195979118347, + "num_tokens": 6778438.0, + "step": 936 + }, + { + "epoch": 0.9551478083588175, + "grad_norm": 7.379710674285889, + "learning_rate": 1e-06, + "loss": 0.3747, + "mean_token_accuracy": 0.844245195388794, + "num_tokens": 6785514.0, + "step": 937 + }, + { + "epoch": 0.9561671763506626, + "grad_norm": 5.246002197265625, + "learning_rate": 1e-06, + "loss": 0.3952, + "mean_token_accuracy": 0.8406311869621277, + "num_tokens": 6793081.0, + "step": 938 + }, + { + "epoch": 0.9571865443425076, + "grad_norm": 4.832677841186523, + "learning_rate": 1e-06, + "loss": 0.3672, + "mean_token_accuracy": 0.8578288555145264, + "num_tokens": 6800353.0, + "step": 939 + }, + { + "epoch": 0.9582059123343527, + "grad_norm": 4.992971897125244, + "learning_rate": 1e-06, + "loss": 0.3629, + "mean_token_accuracy": 0.8567060232162476, + "num_tokens": 6808707.0, + "step": 940 + }, + { + "epoch": 0.9592252803261978, + "grad_norm": 4.774245738983154, + "learning_rate": 1e-06, + "loss": 0.3882, + "mean_token_accuracy": 0.8463152647018433, + "num_tokens": 6816814.0, + "step": 941 + }, + { + "epoch": 0.9602446483180428, + "grad_norm": 5.2275390625, + "learning_rate": 1e-06, + "loss": 0.4571, + "mean_token_accuracy": 0.819718599319458, + "num_tokens": 6825410.0, + "step": 942 + }, + { + "epoch": 0.9612640163098879, + "grad_norm": 5.805734634399414, + "learning_rate": 1e-06, + "loss": 0.4223, + "mean_token_accuracy": 0.8245697021484375, + "num_tokens": 6832368.0, + "step": 943 + }, + { + "epoch": 0.9622833843017329, + "grad_norm": 5.768489360809326, + "learning_rate": 1e-06, + "loss": 0.415, + "mean_token_accuracy": 0.8360896110534668, + "num_tokens": 6839148.0, + "step": 944 + }, + { + "epoch": 0.963302752293578, + "grad_norm": 6.296600818634033, + "learning_rate": 1e-06, + "loss": 0.4269, + "mean_token_accuracy": 0.8315436244010925, + "num_tokens": 6845173.0, + "step": 945 + }, + { + "epoch": 0.9643221202854231, + "grad_norm": 5.537043571472168, + "learning_rate": 1e-06, + "loss": 0.4045, + "mean_token_accuracy": 0.8402498960494995, + "num_tokens": 6852499.0, + "step": 946 + }, + { + "epoch": 0.9653414882772681, + "grad_norm": 4.9958391189575195, + "learning_rate": 1e-06, + "loss": 0.3838, + "mean_token_accuracy": 0.844931960105896, + "num_tokens": 6860848.0, + "step": 947 + }, + { + "epoch": 0.9663608562691132, + "grad_norm": 4.988476753234863, + "learning_rate": 1e-06, + "loss": 0.4019, + "mean_token_accuracy": 0.8398033380508423, + "num_tokens": 6868829.0, + "step": 948 + }, + { + "epoch": 0.9673802242609582, + "grad_norm": 5.393559455871582, + "learning_rate": 1e-06, + "loss": 0.3933, + "mean_token_accuracy": 0.845100998878479, + "num_tokens": 6875842.0, + "step": 949 + }, + { + "epoch": 0.9683995922528033, + "grad_norm": 5.8898115158081055, + "learning_rate": 1e-06, + "loss": 0.3977, + "mean_token_accuracy": 0.8462821245193481, + "num_tokens": 6881672.0, + "step": 950 + }, + { + "epoch": 0.9694189602446484, + "grad_norm": 4.866647243499756, + "learning_rate": 1e-06, + "loss": 0.3535, + "mean_token_accuracy": 0.8557361364364624, + "num_tokens": 6890150.0, + "step": 951 + }, + { + "epoch": 0.9704383282364933, + "grad_norm": 5.921017646789551, + "learning_rate": 1e-06, + "loss": 0.3897, + "mean_token_accuracy": 0.8458033204078674, + "num_tokens": 6896224.0, + "step": 952 + }, + { + "epoch": 0.9714576962283384, + "grad_norm": 4.729032039642334, + "learning_rate": 1e-06, + "loss": 0.3633, + "mean_token_accuracy": 0.8551243543624878, + "num_tokens": 6903542.0, + "step": 953 + }, + { + "epoch": 0.9724770642201835, + "grad_norm": 5.621241092681885, + "learning_rate": 1e-06, + "loss": 0.4251, + "mean_token_accuracy": 0.8339982032775879, + "num_tokens": 6910363.0, + "step": 954 + }, + { + "epoch": 0.9734964322120285, + "grad_norm": 4.206593036651611, + "learning_rate": 1e-06, + "loss": 0.3303, + "mean_token_accuracy": 0.8578614592552185, + "num_tokens": 6919460.0, + "step": 955 + }, + { + "epoch": 0.9745158002038736, + "grad_norm": 4.761197566986084, + "learning_rate": 1e-06, + "loss": 0.3928, + "mean_token_accuracy": 0.8438119888305664, + "num_tokens": 6927987.0, + "step": 956 + }, + { + "epoch": 0.9755351681957186, + "grad_norm": 4.981130123138428, + "learning_rate": 1e-06, + "loss": 0.4264, + "mean_token_accuracy": 0.8356783390045166, + "num_tokens": 6936054.0, + "step": 957 + }, + { + "epoch": 0.9765545361875637, + "grad_norm": 5.769023895263672, + "learning_rate": 1e-06, + "loss": 0.4272, + "mean_token_accuracy": 0.836275577545166, + "num_tokens": 6943485.0, + "step": 958 + }, + { + "epoch": 0.9775739041794088, + "grad_norm": 5.205333232879639, + "learning_rate": 1e-06, + "loss": 0.409, + "mean_token_accuracy": 0.836747944355011, + "num_tokens": 6950703.0, + "step": 959 + }, + { + "epoch": 0.9785932721712538, + "grad_norm": 6.094625949859619, + "learning_rate": 1e-06, + "loss": 0.4251, + "mean_token_accuracy": 0.837172269821167, + "num_tokens": 6956312.0, + "step": 960 + }, + { + "epoch": 0.9796126401630989, + "grad_norm": 5.763320446014404, + "learning_rate": 1e-06, + "loss": 0.401, + "mean_token_accuracy": 0.8380165696144104, + "num_tokens": 6962575.0, + "step": 961 + }, + { + "epoch": 0.9806320081549439, + "grad_norm": 5.624131679534912, + "learning_rate": 1e-06, + "loss": 0.3741, + "mean_token_accuracy": 0.8490338325500488, + "num_tokens": 6968569.0, + "step": 962 + }, + { + "epoch": 0.981651376146789, + "grad_norm": 5.648481369018555, + "learning_rate": 1e-06, + "loss": 0.3762, + "mean_token_accuracy": 0.8525385856628418, + "num_tokens": 6975400.0, + "step": 963 + }, + { + "epoch": 0.9826707441386341, + "grad_norm": 5.409857273101807, + "learning_rate": 1e-06, + "loss": 0.3846, + "mean_token_accuracy": 0.8437213897705078, + "num_tokens": 6981596.0, + "step": 964 + }, + { + "epoch": 0.9836901121304791, + "grad_norm": 5.419344425201416, + "learning_rate": 1e-06, + "loss": 0.4113, + "mean_token_accuracy": 0.8386217355728149, + "num_tokens": 6988992.0, + "step": 965 + }, + { + "epoch": 0.9847094801223242, + "grad_norm": 4.956860542297363, + "learning_rate": 1e-06, + "loss": 0.4075, + "mean_token_accuracy": 0.8442258238792419, + "num_tokens": 6997796.0, + "step": 966 + }, + { + "epoch": 0.9857288481141692, + "grad_norm": 5.536783218383789, + "learning_rate": 1e-06, + "loss": 0.4122, + "mean_token_accuracy": 0.8364111185073853, + "num_tokens": 7005539.0, + "step": 967 + }, + { + "epoch": 0.9867482161060143, + "grad_norm": 5.34040641784668, + "learning_rate": 1e-06, + "loss": 0.4194, + "mean_token_accuracy": 0.8382949829101562, + "num_tokens": 7011759.0, + "step": 968 + }, + { + "epoch": 0.9877675840978594, + "grad_norm": 5.367779731750488, + "learning_rate": 1e-06, + "loss": 0.4037, + "mean_token_accuracy": 0.8421748876571655, + "num_tokens": 7019694.0, + "step": 969 + }, + { + "epoch": 0.9887869520897044, + "grad_norm": 4.366479873657227, + "learning_rate": 1e-06, + "loss": 0.3473, + "mean_token_accuracy": 0.8603573441505432, + "num_tokens": 7028049.0, + "step": 970 + }, + { + "epoch": 0.9898063200815495, + "grad_norm": 5.234898567199707, + "learning_rate": 1e-06, + "loss": 0.3999, + "mean_token_accuracy": 0.8418267965316772, + "num_tokens": 7035318.0, + "step": 971 + }, + { + "epoch": 0.9908256880733946, + "grad_norm": 7.562460422515869, + "learning_rate": 1e-06, + "loss": 0.4252, + "mean_token_accuracy": 0.8331997394561768, + "num_tokens": 7040424.0, + "step": 972 + }, + { + "epoch": 0.9918450560652395, + "grad_norm": 5.684411525726318, + "learning_rate": 1e-06, + "loss": 0.4016, + "mean_token_accuracy": 0.8406955599784851, + "num_tokens": 7047066.0, + "step": 973 + }, + { + "epoch": 0.9928644240570846, + "grad_norm": 4.991189956665039, + "learning_rate": 1e-06, + "loss": 0.3801, + "mean_token_accuracy": 0.8518627882003784, + "num_tokens": 7056323.0, + "step": 974 + }, + { + "epoch": 0.9938837920489296, + "grad_norm": 5.465914249420166, + "learning_rate": 1e-06, + "loss": 0.4127, + "mean_token_accuracy": 0.8335734009742737, + "num_tokens": 7063404.0, + "step": 975 + }, + { + "epoch": 0.9949031600407747, + "grad_norm": 4.788977146148682, + "learning_rate": 1e-06, + "loss": 0.4064, + "mean_token_accuracy": 0.8408795595169067, + "num_tokens": 7071640.0, + "step": 976 + }, + { + "epoch": 0.9959225280326198, + "grad_norm": 6.170888900756836, + "learning_rate": 1e-06, + "loss": 0.4097, + "mean_token_accuracy": 0.8420400619506836, + "num_tokens": 7077901.0, + "step": 977 + }, + { + "epoch": 0.9969418960244648, + "grad_norm": 5.020096302032471, + "learning_rate": 1e-06, + "loss": 0.3932, + "mean_token_accuracy": 0.8402166962623596, + "num_tokens": 7085628.0, + "step": 978 + }, + { + "epoch": 0.9979612640163099, + "grad_norm": 6.717462062835693, + "learning_rate": 1e-06, + "loss": 0.4594, + "mean_token_accuracy": 0.8270029425621033, + "num_tokens": 7091705.0, + "step": 979 + }, + { + "epoch": 0.9989806320081549, + "grad_norm": 5.561420917510986, + "learning_rate": 1e-06, + "loss": 0.3949, + "mean_token_accuracy": 0.8493162393569946, + "num_tokens": 7099477.0, + "step": 980 + }, + { + "epoch": 1.0, + "grad_norm": 5.343495845794678, + "learning_rate": 1e-06, + "loss": 0.3933, + "mean_token_accuracy": 0.8407586812973022, + "num_tokens": 7107684.0, + "step": 981 + }, + { + "epoch": 1.001019367991845, + "grad_norm": 4.809633731842041, + "learning_rate": 1e-06, + "loss": 0.3278, + "mean_token_accuracy": 0.8777881264686584, + "num_tokens": 7115394.0, + "step": 982 + }, + { + "epoch": 1.0020387359836902, + "grad_norm": 5.225322723388672, + "learning_rate": 1e-06, + "loss": 0.3384, + "mean_token_accuracy": 0.8670807480812073, + "num_tokens": 7121690.0, + "step": 983 + }, + { + "epoch": 1.003058103975535, + "grad_norm": 4.934613227844238, + "learning_rate": 1e-06, + "loss": 0.3294, + "mean_token_accuracy": 0.8672311305999756, + "num_tokens": 7128685.0, + "step": 984 + }, + { + "epoch": 1.0040774719673802, + "grad_norm": 4.601254463195801, + "learning_rate": 1e-06, + "loss": 0.3, + "mean_token_accuracy": 0.8839112520217896, + "num_tokens": 7136133.0, + "step": 985 + }, + { + "epoch": 1.0050968399592253, + "grad_norm": 4.677940368652344, + "learning_rate": 1e-06, + "loss": 0.3039, + "mean_token_accuracy": 0.8812642097473145, + "num_tokens": 7144213.0, + "step": 986 + }, + { + "epoch": 1.0061162079510704, + "grad_norm": 4.989991188049316, + "learning_rate": 1e-06, + "loss": 0.3144, + "mean_token_accuracy": 0.8740074038505554, + "num_tokens": 7151728.0, + "step": 987 + }, + { + "epoch": 1.0071355759429155, + "grad_norm": 5.47297477722168, + "learning_rate": 1e-06, + "loss": 0.3145, + "mean_token_accuracy": 0.8820443749427795, + "num_tokens": 7158559.0, + "step": 988 + }, + { + "epoch": 1.0081549439347604, + "grad_norm": 5.976856231689453, + "learning_rate": 1e-06, + "loss": 0.3298, + "mean_token_accuracy": 0.8712736368179321, + "num_tokens": 7164651.0, + "step": 989 + }, + { + "epoch": 1.0091743119266054, + "grad_norm": 5.367843151092529, + "learning_rate": 1e-06, + "loss": 0.3035, + "mean_token_accuracy": 0.8837795257568359, + "num_tokens": 7172091.0, + "step": 990 + }, + { + "epoch": 1.0101936799184505, + "grad_norm": 5.4621429443359375, + "learning_rate": 1e-06, + "loss": 0.2966, + "mean_token_accuracy": 0.8823059797286987, + "num_tokens": 7178842.0, + "step": 991 + }, + { + "epoch": 1.0112130479102956, + "grad_norm": 6.558964252471924, + "learning_rate": 1e-06, + "loss": 0.3461, + "mean_token_accuracy": 0.8666529655456543, + "num_tokens": 7185352.0, + "step": 992 + }, + { + "epoch": 1.0122324159021407, + "grad_norm": 5.794253826141357, + "learning_rate": 1e-06, + "loss": 0.301, + "mean_token_accuracy": 0.8876056671142578, + "num_tokens": 7192765.0, + "step": 993 + }, + { + "epoch": 1.0132517838939856, + "grad_norm": 6.314942836761475, + "learning_rate": 1e-06, + "loss": 0.3109, + "mean_token_accuracy": 0.876176118850708, + "num_tokens": 7200020.0, + "step": 994 + }, + { + "epoch": 1.0142711518858307, + "grad_norm": 6.809550762176514, + "learning_rate": 1e-06, + "loss": 0.3429, + "mean_token_accuracy": 0.8675749897956848, + "num_tokens": 7206880.0, + "step": 995 + }, + { + "epoch": 1.0152905198776758, + "grad_norm": 5.442190170288086, + "learning_rate": 1e-06, + "loss": 0.3187, + "mean_token_accuracy": 0.8717448711395264, + "num_tokens": 7214792.0, + "step": 996 + }, + { + "epoch": 1.016309887869521, + "grad_norm": 5.585460186004639, + "learning_rate": 1e-06, + "loss": 0.3005, + "mean_token_accuracy": 0.8822904825210571, + "num_tokens": 7220977.0, + "step": 997 + }, + { + "epoch": 1.017329255861366, + "grad_norm": 5.197250843048096, + "learning_rate": 1e-06, + "loss": 0.3238, + "mean_token_accuracy": 0.8705838918685913, + "num_tokens": 7228290.0, + "step": 998 + }, + { + "epoch": 1.018348623853211, + "grad_norm": 4.895814418792725, + "learning_rate": 1e-06, + "loss": 0.2647, + "mean_token_accuracy": 0.8967143297195435, + "num_tokens": 7235622.0, + "step": 999 + }, + { + "epoch": 1.019367991845056, + "grad_norm": 4.937918663024902, + "learning_rate": 1e-06, + "loss": 0.2913, + "mean_token_accuracy": 0.8877084851264954, + "num_tokens": 7243680.0, + "step": 1000 + }, + { + "epoch": 1.020387359836901, + "grad_norm": 4.450643062591553, + "learning_rate": 1e-06, + "loss": 0.2538, + "mean_token_accuracy": 0.9005272388458252, + "num_tokens": 7251667.0, + "step": 1001 + }, + { + "epoch": 1.0214067278287462, + "grad_norm": 4.698352336883545, + "learning_rate": 1e-06, + "loss": 0.2713, + "mean_token_accuracy": 0.8920304775238037, + "num_tokens": 7259958.0, + "step": 1002 + }, + { + "epoch": 1.0224260958205913, + "grad_norm": 5.438255310058594, + "learning_rate": 1e-06, + "loss": 0.2994, + "mean_token_accuracy": 0.8776328563690186, + "num_tokens": 7266843.0, + "step": 1003 + }, + { + "epoch": 1.0234454638124364, + "grad_norm": 5.070213794708252, + "learning_rate": 1e-06, + "loss": 0.3045, + "mean_token_accuracy": 0.8814657330513, + "num_tokens": 7274038.0, + "step": 1004 + }, + { + "epoch": 1.0244648318042813, + "grad_norm": 6.410051345825195, + "learning_rate": 1e-06, + "loss": 0.3017, + "mean_token_accuracy": 0.877382755279541, + "num_tokens": 7279123.0, + "step": 1005 + }, + { + "epoch": 1.0254841997961264, + "grad_norm": 4.843079090118408, + "learning_rate": 1e-06, + "loss": 0.3293, + "mean_token_accuracy": 0.8820123672485352, + "num_tokens": 7287377.0, + "step": 1006 + }, + { + "epoch": 1.0265035677879715, + "grad_norm": 6.856749057769775, + "learning_rate": 1e-06, + "loss": 0.3499, + "mean_token_accuracy": 0.8609369993209839, + "num_tokens": 7292696.0, + "step": 1007 + }, + { + "epoch": 1.0275229357798166, + "grad_norm": 6.313785552978516, + "learning_rate": 1e-06, + "loss": 0.3458, + "mean_token_accuracy": 0.86699378490448, + "num_tokens": 7299146.0, + "step": 1008 + }, + { + "epoch": 1.0285423037716617, + "grad_norm": 4.461640357971191, + "learning_rate": 1e-06, + "loss": 0.2822, + "mean_token_accuracy": 0.8903945088386536, + "num_tokens": 7308845.0, + "step": 1009 + }, + { + "epoch": 1.0295616717635065, + "grad_norm": 5.746631145477295, + "learning_rate": 1e-06, + "loss": 0.2985, + "mean_token_accuracy": 0.8783943057060242, + "num_tokens": 7316473.0, + "step": 1010 + }, + { + "epoch": 1.0305810397553516, + "grad_norm": 5.473225116729736, + "learning_rate": 1e-06, + "loss": 0.3187, + "mean_token_accuracy": 0.8755156993865967, + "num_tokens": 7323964.0, + "step": 1011 + }, + { + "epoch": 1.0316004077471967, + "grad_norm": 5.864831924438477, + "learning_rate": 1e-06, + "loss": 0.3394, + "mean_token_accuracy": 0.8657529950141907, + "num_tokens": 7330894.0, + "step": 1012 + }, + { + "epoch": 1.0326197757390418, + "grad_norm": 5.464033126831055, + "learning_rate": 1e-06, + "loss": 0.3216, + "mean_token_accuracy": 0.8727334141731262, + "num_tokens": 7337182.0, + "step": 1013 + }, + { + "epoch": 1.033639143730887, + "grad_norm": 4.994362831115723, + "learning_rate": 1e-06, + "loss": 0.3025, + "mean_token_accuracy": 0.87621009349823, + "num_tokens": 7344566.0, + "step": 1014 + }, + { + "epoch": 1.0346585117227318, + "grad_norm": 4.575716972351074, + "learning_rate": 1e-06, + "loss": 0.2928, + "mean_token_accuracy": 0.8836846351623535, + "num_tokens": 7353402.0, + "step": 1015 + }, + { + "epoch": 1.035677879714577, + "grad_norm": 4.005095958709717, + "learning_rate": 1e-06, + "loss": 0.2522, + "mean_token_accuracy": 0.9016960263252258, + "num_tokens": 7361928.0, + "step": 1016 + }, + { + "epoch": 1.036697247706422, + "grad_norm": 5.906186103820801, + "learning_rate": 1e-06, + "loss": 0.3057, + "mean_token_accuracy": 0.882417619228363, + "num_tokens": 7368541.0, + "step": 1017 + }, + { + "epoch": 1.0377166156982671, + "grad_norm": 5.134413242340088, + "learning_rate": 1e-06, + "loss": 0.3188, + "mean_token_accuracy": 0.8759173154830933, + "num_tokens": 7375221.0, + "step": 1018 + }, + { + "epoch": 1.0387359836901122, + "grad_norm": 5.336219310760498, + "learning_rate": 1e-06, + "loss": 0.2975, + "mean_token_accuracy": 0.8862130045890808, + "num_tokens": 7382023.0, + "step": 1019 + }, + { + "epoch": 1.039755351681957, + "grad_norm": 5.172081470489502, + "learning_rate": 1e-06, + "loss": 0.3086, + "mean_token_accuracy": 0.8815064430236816, + "num_tokens": 7390105.0, + "step": 1020 + }, + { + "epoch": 1.0407747196738022, + "grad_norm": 4.3155741691589355, + "learning_rate": 1e-06, + "loss": 0.283, + "mean_token_accuracy": 0.8771250247955322, + "num_tokens": 7399378.0, + "step": 1021 + }, + { + "epoch": 1.0417940876656473, + "grad_norm": 4.882828235626221, + "learning_rate": 1e-06, + "loss": 0.2777, + "mean_token_accuracy": 0.8765301704406738, + "num_tokens": 7406266.0, + "step": 1022 + }, + { + "epoch": 1.0428134556574924, + "grad_norm": 5.121819496154785, + "learning_rate": 1e-06, + "loss": 0.2862, + "mean_token_accuracy": 0.8858315944671631, + "num_tokens": 7413230.0, + "step": 1023 + }, + { + "epoch": 1.0438328236493375, + "grad_norm": 6.0568766593933105, + "learning_rate": 1e-06, + "loss": 0.3342, + "mean_token_accuracy": 0.8654904365539551, + "num_tokens": 7419902.0, + "step": 1024 + }, + { + "epoch": 1.0448521916411824, + "grad_norm": 5.409205436706543, + "learning_rate": 1e-06, + "loss": 0.3011, + "mean_token_accuracy": 0.879853367805481, + "num_tokens": 7426869.0, + "step": 1025 + }, + { + "epoch": 1.0458715596330275, + "grad_norm": 5.642271041870117, + "learning_rate": 1e-06, + "loss": 0.2778, + "mean_token_accuracy": 0.8876335620880127, + "num_tokens": 7433908.0, + "step": 1026 + }, + { + "epoch": 1.0468909276248726, + "grad_norm": 5.5788679122924805, + "learning_rate": 1e-06, + "loss": 0.2942, + "mean_token_accuracy": 0.8768489360809326, + "num_tokens": 7441032.0, + "step": 1027 + }, + { + "epoch": 1.0479102956167177, + "grad_norm": 6.108129024505615, + "learning_rate": 1e-06, + "loss": 0.3438, + "mean_token_accuracy": 0.8633602261543274, + "num_tokens": 7447533.0, + "step": 1028 + }, + { + "epoch": 1.0489296636085628, + "grad_norm": 5.8037872314453125, + "learning_rate": 1e-06, + "loss": 0.3024, + "mean_token_accuracy": 0.8839191198348999, + "num_tokens": 7454681.0, + "step": 1029 + }, + { + "epoch": 1.0499490316004076, + "grad_norm": 5.81805944442749, + "learning_rate": 1e-06, + "loss": 0.3381, + "mean_token_accuracy": 0.8673350811004639, + "num_tokens": 7461995.0, + "step": 1030 + }, + { + "epoch": 1.0509683995922527, + "grad_norm": 4.93458890914917, + "learning_rate": 1e-06, + "loss": 0.2577, + "mean_token_accuracy": 0.8996323347091675, + "num_tokens": 7469474.0, + "step": 1031 + }, + { + "epoch": 1.0519877675840978, + "grad_norm": 4.653446674346924, + "learning_rate": 1e-06, + "loss": 0.2873, + "mean_token_accuracy": 0.8873215913772583, + "num_tokens": 7478209.0, + "step": 1032 + }, + { + "epoch": 1.053007135575943, + "grad_norm": 4.788922309875488, + "learning_rate": 1e-06, + "loss": 0.29, + "mean_token_accuracy": 0.884804368019104, + "num_tokens": 7485784.0, + "step": 1033 + }, + { + "epoch": 1.054026503567788, + "grad_norm": 5.290215969085693, + "learning_rate": 1e-06, + "loss": 0.3143, + "mean_token_accuracy": 0.8784863948822021, + "num_tokens": 7492817.0, + "step": 1034 + }, + { + "epoch": 1.0550458715596331, + "grad_norm": 5.54358434677124, + "learning_rate": 1e-06, + "loss": 0.2996, + "mean_token_accuracy": 0.8867706060409546, + "num_tokens": 7499726.0, + "step": 1035 + }, + { + "epoch": 1.056065239551478, + "grad_norm": 5.712363243103027, + "learning_rate": 1e-06, + "loss": 0.3307, + "mean_token_accuracy": 0.871013879776001, + "num_tokens": 7506196.0, + "step": 1036 + }, + { + "epoch": 1.0570846075433231, + "grad_norm": 5.126567363739014, + "learning_rate": 1e-06, + "loss": 0.3017, + "mean_token_accuracy": 0.8811429738998413, + "num_tokens": 7512935.0, + "step": 1037 + }, + { + "epoch": 1.0581039755351682, + "grad_norm": 5.075089454650879, + "learning_rate": 1e-06, + "loss": 0.3191, + "mean_token_accuracy": 0.8731293678283691, + "num_tokens": 7520425.0, + "step": 1038 + }, + { + "epoch": 1.0591233435270133, + "grad_norm": 4.516158580780029, + "learning_rate": 1e-06, + "loss": 0.2497, + "mean_token_accuracy": 0.8950295448303223, + "num_tokens": 7530719.0, + "step": 1039 + }, + { + "epoch": 1.0601427115188584, + "grad_norm": 6.1354851722717285, + "learning_rate": 1e-06, + "loss": 0.343, + "mean_token_accuracy": 0.8664819002151489, + "num_tokens": 7536526.0, + "step": 1040 + }, + { + "epoch": 1.0611620795107033, + "grad_norm": 5.4319634437561035, + "learning_rate": 1e-06, + "loss": 0.2994, + "mean_token_accuracy": 0.8828256130218506, + "num_tokens": 7543706.0, + "step": 1041 + }, + { + "epoch": 1.0621814475025484, + "grad_norm": 5.645985126495361, + "learning_rate": 1e-06, + "loss": 0.2965, + "mean_token_accuracy": 0.8809824585914612, + "num_tokens": 7550099.0, + "step": 1042 + }, + { + "epoch": 1.0632008154943935, + "grad_norm": 5.573832035064697, + "learning_rate": 1e-06, + "loss": 0.2892, + "mean_token_accuracy": 0.8892544507980347, + "num_tokens": 7556454.0, + "step": 1043 + }, + { + "epoch": 1.0642201834862386, + "grad_norm": 6.490408897399902, + "learning_rate": 1e-06, + "loss": 0.3284, + "mean_token_accuracy": 0.8660033941268921, + "num_tokens": 7561967.0, + "step": 1044 + }, + { + "epoch": 1.0652395514780837, + "grad_norm": 6.401559829711914, + "learning_rate": 1e-06, + "loss": 0.3355, + "mean_token_accuracy": 0.8650749921798706, + "num_tokens": 7568398.0, + "step": 1045 + }, + { + "epoch": 1.0662589194699286, + "grad_norm": 5.785712242126465, + "learning_rate": 1e-06, + "loss": 0.2979, + "mean_token_accuracy": 0.876246452331543, + "num_tokens": 7575659.0, + "step": 1046 + }, + { + "epoch": 1.0672782874617737, + "grad_norm": 5.075110912322998, + "learning_rate": 1e-06, + "loss": 0.2613, + "mean_token_accuracy": 0.8868962526321411, + "num_tokens": 7582971.0, + "step": 1047 + }, + { + "epoch": 1.0682976554536188, + "grad_norm": 6.003932476043701, + "learning_rate": 1e-06, + "loss": 0.3144, + "mean_token_accuracy": 0.8796887397766113, + "num_tokens": 7588856.0, + "step": 1048 + }, + { + "epoch": 1.0693170234454639, + "grad_norm": 5.176914215087891, + "learning_rate": 1e-06, + "loss": 0.3185, + "mean_token_accuracy": 0.8742461800575256, + "num_tokens": 7596356.0, + "step": 1049 + }, + { + "epoch": 1.070336391437309, + "grad_norm": 6.108354091644287, + "learning_rate": 1e-06, + "loss": 0.3427, + "mean_token_accuracy": 0.8585034608840942, + "num_tokens": 7603506.0, + "step": 1050 + }, + { + "epoch": 1.0713557594291538, + "grad_norm": 5.0289106369018555, + "learning_rate": 1e-06, + "loss": 0.2724, + "mean_token_accuracy": 0.8948214054107666, + "num_tokens": 7611066.0, + "step": 1051 + }, + { + "epoch": 1.072375127420999, + "grad_norm": 5.274388313293457, + "learning_rate": 1e-06, + "loss": 0.2781, + "mean_token_accuracy": 0.8954247832298279, + "num_tokens": 7618638.0, + "step": 1052 + }, + { + "epoch": 1.073394495412844, + "grad_norm": 4.772217750549316, + "learning_rate": 1e-06, + "loss": 0.3171, + "mean_token_accuracy": 0.8730885982513428, + "num_tokens": 7626005.0, + "step": 1053 + }, + { + "epoch": 1.0744138634046891, + "grad_norm": 5.408519268035889, + "learning_rate": 1e-06, + "loss": 0.3005, + "mean_token_accuracy": 0.8818575739860535, + "num_tokens": 7633303.0, + "step": 1054 + }, + { + "epoch": 1.0754332313965342, + "grad_norm": 5.587804794311523, + "learning_rate": 1e-06, + "loss": 0.2986, + "mean_token_accuracy": 0.8872005343437195, + "num_tokens": 7639661.0, + "step": 1055 + }, + { + "epoch": 1.0764525993883791, + "grad_norm": 7.019844055175781, + "learning_rate": 1e-06, + "loss": 0.3585, + "mean_token_accuracy": 0.8541271686553955, + "num_tokens": 7644999.0, + "step": 1056 + }, + { + "epoch": 1.0774719673802242, + "grad_norm": 5.849760055541992, + "learning_rate": 1e-06, + "loss": 0.3329, + "mean_token_accuracy": 0.8710834383964539, + "num_tokens": 7651775.0, + "step": 1057 + }, + { + "epoch": 1.0784913353720693, + "grad_norm": 6.278073310852051, + "learning_rate": 1e-06, + "loss": 0.2987, + "mean_token_accuracy": 0.8811357617378235, + "num_tokens": 7658706.0, + "step": 1058 + }, + { + "epoch": 1.0795107033639144, + "grad_norm": 5.427402019500732, + "learning_rate": 1e-06, + "loss": 0.2995, + "mean_token_accuracy": 0.8815218210220337, + "num_tokens": 7665494.0, + "step": 1059 + }, + { + "epoch": 1.0805300713557595, + "grad_norm": 5.098483562469482, + "learning_rate": 1e-06, + "loss": 0.2818, + "mean_token_accuracy": 0.8920868635177612, + "num_tokens": 7672448.0, + "step": 1060 + }, + { + "epoch": 1.0815494393476044, + "grad_norm": 4.870551109313965, + "learning_rate": 1e-06, + "loss": 0.268, + "mean_token_accuracy": 0.8930174112319946, + "num_tokens": 7679904.0, + "step": 1061 + }, + { + "epoch": 1.0825688073394495, + "grad_norm": 6.88853645324707, + "learning_rate": 1e-06, + "loss": 0.3303, + "mean_token_accuracy": 0.8610320687294006, + "num_tokens": 7685683.0, + "step": 1062 + }, + { + "epoch": 1.0835881753312946, + "grad_norm": 6.126386642456055, + "learning_rate": 1e-06, + "loss": 0.3094, + "mean_token_accuracy": 0.8814899325370789, + "num_tokens": 7692357.0, + "step": 1063 + }, + { + "epoch": 1.0846075433231397, + "grad_norm": 5.731656074523926, + "learning_rate": 1e-06, + "loss": 0.3003, + "mean_token_accuracy": 0.8787443041801453, + "num_tokens": 7699757.0, + "step": 1064 + }, + { + "epoch": 1.0856269113149848, + "grad_norm": 4.485128402709961, + "learning_rate": 1e-06, + "loss": 0.302, + "mean_token_accuracy": 0.8819250464439392, + "num_tokens": 7708121.0, + "step": 1065 + }, + { + "epoch": 1.0866462793068297, + "grad_norm": 6.009196758270264, + "learning_rate": 1e-06, + "loss": 0.2952, + "mean_token_accuracy": 0.8777909278869629, + "num_tokens": 7713859.0, + "step": 1066 + }, + { + "epoch": 1.0876656472986748, + "grad_norm": 4.550666332244873, + "learning_rate": 1e-06, + "loss": 0.2705, + "mean_token_accuracy": 0.8911813497543335, + "num_tokens": 7721859.0, + "step": 1067 + }, + { + "epoch": 1.0886850152905199, + "grad_norm": 5.5806355476379395, + "learning_rate": 1e-06, + "loss": 0.3045, + "mean_token_accuracy": 0.8851617574691772, + "num_tokens": 7728474.0, + "step": 1068 + }, + { + "epoch": 1.089704383282365, + "grad_norm": 5.8889923095703125, + "learning_rate": 1e-06, + "loss": 0.3239, + "mean_token_accuracy": 0.8776732087135315, + "num_tokens": 7736349.0, + "step": 1069 + }, + { + "epoch": 1.09072375127421, + "grad_norm": 6.844680309295654, + "learning_rate": 1e-06, + "loss": 0.3467, + "mean_token_accuracy": 0.8662342429161072, + "num_tokens": 7742295.0, + "step": 1070 + }, + { + "epoch": 1.091743119266055, + "grad_norm": 5.8229756355285645, + "learning_rate": 1e-06, + "loss": 0.3257, + "mean_token_accuracy": 0.8794769644737244, + "num_tokens": 7749785.0, + "step": 1071 + }, + { + "epoch": 1.0927624872579, + "grad_norm": 5.780683994293213, + "learning_rate": 1e-06, + "loss": 0.3029, + "mean_token_accuracy": 0.8794364929199219, + "num_tokens": 7756679.0, + "step": 1072 + }, + { + "epoch": 1.0937818552497451, + "grad_norm": 5.908757209777832, + "learning_rate": 1e-06, + "loss": 0.3283, + "mean_token_accuracy": 0.8770340085029602, + "num_tokens": 7764345.0, + "step": 1073 + }, + { + "epoch": 1.0948012232415902, + "grad_norm": 5.634097099304199, + "learning_rate": 1e-06, + "loss": 0.2995, + "mean_token_accuracy": 0.8885225057601929, + "num_tokens": 7771519.0, + "step": 1074 + }, + { + "epoch": 1.0958205912334353, + "grad_norm": 5.942012786865234, + "learning_rate": 1e-06, + "loss": 0.2956, + "mean_token_accuracy": 0.8831430673599243, + "num_tokens": 7777909.0, + "step": 1075 + }, + { + "epoch": 1.0968399592252802, + "grad_norm": 4.695597171783447, + "learning_rate": 1e-06, + "loss": 0.2811, + "mean_token_accuracy": 0.8846499919891357, + "num_tokens": 7787261.0, + "step": 1076 + }, + { + "epoch": 1.0978593272171253, + "grad_norm": 6.425548553466797, + "learning_rate": 1e-06, + "loss": 0.312, + "mean_token_accuracy": 0.8792320489883423, + "num_tokens": 7793621.0, + "step": 1077 + }, + { + "epoch": 1.0988786952089704, + "grad_norm": 5.317170143127441, + "learning_rate": 1e-06, + "loss": 0.3003, + "mean_token_accuracy": 0.8764793276786804, + "num_tokens": 7800165.0, + "step": 1078 + }, + { + "epoch": 1.0998980632008155, + "grad_norm": 4.836089134216309, + "learning_rate": 1e-06, + "loss": 0.3021, + "mean_token_accuracy": 0.8799510598182678, + "num_tokens": 7807794.0, + "step": 1079 + }, + { + "epoch": 1.1009174311926606, + "grad_norm": 6.510807991027832, + "learning_rate": 1e-06, + "loss": 0.3051, + "mean_token_accuracy": 0.8835996985435486, + "num_tokens": 7813716.0, + "step": 1080 + }, + { + "epoch": 1.1019367991845055, + "grad_norm": 6.03483247756958, + "learning_rate": 1e-06, + "loss": 0.3174, + "mean_token_accuracy": 0.8726514577865601, + "num_tokens": 7819741.0, + "step": 1081 + }, + { + "epoch": 1.1029561671763506, + "grad_norm": 5.9372453689575195, + "learning_rate": 1e-06, + "loss": 0.3069, + "mean_token_accuracy": 0.8812574148178101, + "num_tokens": 7825987.0, + "step": 1082 + }, + { + "epoch": 1.1039755351681957, + "grad_norm": 5.081005573272705, + "learning_rate": 1e-06, + "loss": 0.2807, + "mean_token_accuracy": 0.8843290209770203, + "num_tokens": 7833835.0, + "step": 1083 + }, + { + "epoch": 1.1049949031600408, + "grad_norm": 5.9667205810546875, + "learning_rate": 1e-06, + "loss": 0.3361, + "mean_token_accuracy": 0.8716875314712524, + "num_tokens": 7840676.0, + "step": 1084 + }, + { + "epoch": 1.1060142711518859, + "grad_norm": 5.880688667297363, + "learning_rate": 1e-06, + "loss": 0.2845, + "mean_token_accuracy": 0.8896210193634033, + "num_tokens": 7846415.0, + "step": 1085 + }, + { + "epoch": 1.107033639143731, + "grad_norm": 6.242093086242676, + "learning_rate": 1e-06, + "loss": 0.321, + "mean_token_accuracy": 0.8769176006317139, + "num_tokens": 7852813.0, + "step": 1086 + }, + { + "epoch": 1.1080530071355759, + "grad_norm": 5.645715713500977, + "learning_rate": 1e-06, + "loss": 0.3119, + "mean_token_accuracy": 0.8717231154441833, + "num_tokens": 7859575.0, + "step": 1087 + }, + { + "epoch": 1.109072375127421, + "grad_norm": 5.7526774406433105, + "learning_rate": 1e-06, + "loss": 0.3159, + "mean_token_accuracy": 0.8758944869041443, + "num_tokens": 7865766.0, + "step": 1088 + }, + { + "epoch": 1.110091743119266, + "grad_norm": 5.636137008666992, + "learning_rate": 1e-06, + "loss": 0.3095, + "mean_token_accuracy": 0.8821076154708862, + "num_tokens": 7872476.0, + "step": 1089 + }, + { + "epoch": 1.1111111111111112, + "grad_norm": 4.829869747161865, + "learning_rate": 1e-06, + "loss": 0.3049, + "mean_token_accuracy": 0.8714845776557922, + "num_tokens": 7879564.0, + "step": 1090 + }, + { + "epoch": 1.1121304791029563, + "grad_norm": 5.444368839263916, + "learning_rate": 1e-06, + "loss": 0.3128, + "mean_token_accuracy": 0.8805865049362183, + "num_tokens": 7887865.0, + "step": 1091 + }, + { + "epoch": 1.1131498470948011, + "grad_norm": 5.6001973152160645, + "learning_rate": 1e-06, + "loss": 0.3277, + "mean_token_accuracy": 0.8749939799308777, + "num_tokens": 7895357.0, + "step": 1092 + }, + { + "epoch": 1.1141692150866462, + "grad_norm": 4.819443225860596, + "learning_rate": 1e-06, + "loss": 0.2788, + "mean_token_accuracy": 0.8929961919784546, + "num_tokens": 7903154.0, + "step": 1093 + }, + { + "epoch": 1.1151885830784913, + "grad_norm": 6.0509419441223145, + "learning_rate": 1e-06, + "loss": 0.2876, + "mean_token_accuracy": 0.8897977471351624, + "num_tokens": 7910006.0, + "step": 1094 + }, + { + "epoch": 1.1162079510703364, + "grad_norm": 5.195507526397705, + "learning_rate": 1e-06, + "loss": 0.2917, + "mean_token_accuracy": 0.8864816427230835, + "num_tokens": 7916657.0, + "step": 1095 + }, + { + "epoch": 1.1172273190621815, + "grad_norm": 4.830432415008545, + "learning_rate": 1e-06, + "loss": 0.2855, + "mean_token_accuracy": 0.8901296257972717, + "num_tokens": 7924806.0, + "step": 1096 + }, + { + "epoch": 1.1182466870540264, + "grad_norm": 4.2776079177856445, + "learning_rate": 1e-06, + "loss": 0.2728, + "mean_token_accuracy": 0.8931641578674316, + "num_tokens": 7933114.0, + "step": 1097 + }, + { + "epoch": 1.1192660550458715, + "grad_norm": 5.395074367523193, + "learning_rate": 1e-06, + "loss": 0.3182, + "mean_token_accuracy": 0.8777101039886475, + "num_tokens": 7941944.0, + "step": 1098 + }, + { + "epoch": 1.1202854230377166, + "grad_norm": 5.394089698791504, + "learning_rate": 1e-06, + "loss": 0.3035, + "mean_token_accuracy": 0.880644679069519, + "num_tokens": 7948818.0, + "step": 1099 + }, + { + "epoch": 1.1213047910295617, + "grad_norm": 5.921902179718018, + "learning_rate": 1e-06, + "loss": 0.3164, + "mean_token_accuracy": 0.8793970942497253, + "num_tokens": 7955378.0, + "step": 1100 + }, + { + "epoch": 1.1223241590214068, + "grad_norm": 5.6813788414001465, + "learning_rate": 1e-06, + "loss": 0.2922, + "mean_token_accuracy": 0.8789268136024475, + "num_tokens": 7962073.0, + "step": 1101 + }, + { + "epoch": 1.1233435270132517, + "grad_norm": 4.13659143447876, + "learning_rate": 1e-06, + "loss": 0.241, + "mean_token_accuracy": 0.8986406922340393, + "num_tokens": 7970853.0, + "step": 1102 + }, + { + "epoch": 1.1243628950050968, + "grad_norm": 5.458010673522949, + "learning_rate": 1e-06, + "loss": 0.3019, + "mean_token_accuracy": 0.8903704881668091, + "num_tokens": 7978485.0, + "step": 1103 + }, + { + "epoch": 1.1253822629969419, + "grad_norm": 6.043796062469482, + "learning_rate": 1e-06, + "loss": 0.3296, + "mean_token_accuracy": 0.8774415254592896, + "num_tokens": 7984966.0, + "step": 1104 + }, + { + "epoch": 1.126401630988787, + "grad_norm": 5.891540050506592, + "learning_rate": 1e-06, + "loss": 0.3153, + "mean_token_accuracy": 0.8752431869506836, + "num_tokens": 7991255.0, + "step": 1105 + }, + { + "epoch": 1.127420998980632, + "grad_norm": 5.368124961853027, + "learning_rate": 1e-06, + "loss": 0.2805, + "mean_token_accuracy": 0.8898236155509949, + "num_tokens": 7998364.0, + "step": 1106 + }, + { + "epoch": 1.1284403669724772, + "grad_norm": 4.219986438751221, + "learning_rate": 1e-06, + "loss": 0.2594, + "mean_token_accuracy": 0.8969157934188843, + "num_tokens": 8007157.0, + "step": 1107 + }, + { + "epoch": 1.129459734964322, + "grad_norm": 4.999745845794678, + "learning_rate": 1e-06, + "loss": 0.2988, + "mean_token_accuracy": 0.8830659985542297, + "num_tokens": 8014990.0, + "step": 1108 + }, + { + "epoch": 1.1304791029561672, + "grad_norm": 4.946268558502197, + "learning_rate": 1e-06, + "loss": 0.2932, + "mean_token_accuracy": 0.8875949382781982, + "num_tokens": 8022611.0, + "step": 1109 + }, + { + "epoch": 1.1314984709480123, + "grad_norm": 5.189414024353027, + "learning_rate": 1e-06, + "loss": 0.3008, + "mean_token_accuracy": 0.879704475402832, + "num_tokens": 8029666.0, + "step": 1110 + }, + { + "epoch": 1.1325178389398574, + "grad_norm": 5.1874494552612305, + "learning_rate": 1e-06, + "loss": 0.3076, + "mean_token_accuracy": 0.8800025582313538, + "num_tokens": 8037201.0, + "step": 1111 + }, + { + "epoch": 1.1335372069317025, + "grad_norm": 5.297932147979736, + "learning_rate": 1e-06, + "loss": 0.3161, + "mean_token_accuracy": 0.8803998231887817, + "num_tokens": 8045086.0, + "step": 1112 + }, + { + "epoch": 1.1345565749235473, + "grad_norm": 5.564547061920166, + "learning_rate": 1e-06, + "loss": 0.3178, + "mean_token_accuracy": 0.8762052059173584, + "num_tokens": 8051895.0, + "step": 1113 + }, + { + "epoch": 1.1355759429153924, + "grad_norm": 5.497649669647217, + "learning_rate": 1e-06, + "loss": 0.3169, + "mean_token_accuracy": 0.8749916553497314, + "num_tokens": 8058662.0, + "step": 1114 + }, + { + "epoch": 1.1365953109072375, + "grad_norm": 5.0247931480407715, + "learning_rate": 1e-06, + "loss": 0.3212, + "mean_token_accuracy": 0.8744030594825745, + "num_tokens": 8066747.0, + "step": 1115 + }, + { + "epoch": 1.1376146788990826, + "grad_norm": 5.257003307342529, + "learning_rate": 1e-06, + "loss": 0.2855, + "mean_token_accuracy": 0.8882949352264404, + "num_tokens": 8074479.0, + "step": 1116 + }, + { + "epoch": 1.1386340468909277, + "grad_norm": 4.592959403991699, + "learning_rate": 1e-06, + "loss": 0.2585, + "mean_token_accuracy": 0.8973153829574585, + "num_tokens": 8082985.0, + "step": 1117 + }, + { + "epoch": 1.1396534148827726, + "grad_norm": 5.3409104347229, + "learning_rate": 1e-06, + "loss": 0.2915, + "mean_token_accuracy": 0.8820928335189819, + "num_tokens": 8089776.0, + "step": 1118 + }, + { + "epoch": 1.1406727828746177, + "grad_norm": 4.652228832244873, + "learning_rate": 1e-06, + "loss": 0.2911, + "mean_token_accuracy": 0.8877614736557007, + "num_tokens": 8098622.0, + "step": 1119 + }, + { + "epoch": 1.1416921508664628, + "grad_norm": 5.8328962326049805, + "learning_rate": 1e-06, + "loss": 0.3188, + "mean_token_accuracy": 0.8750899434089661, + "num_tokens": 8106176.0, + "step": 1120 + }, + { + "epoch": 1.142711518858308, + "grad_norm": 4.45005464553833, + "learning_rate": 1e-06, + "loss": 0.2849, + "mean_token_accuracy": 0.8773497343063354, + "num_tokens": 8114323.0, + "step": 1121 + }, + { + "epoch": 1.143730886850153, + "grad_norm": 5.276242733001709, + "learning_rate": 1e-06, + "loss": 0.2625, + "mean_token_accuracy": 0.8931812047958374, + "num_tokens": 8121018.0, + "step": 1122 + }, + { + "epoch": 1.1447502548419979, + "grad_norm": 5.599971771240234, + "learning_rate": 1e-06, + "loss": 0.3404, + "mean_token_accuracy": 0.8663594722747803, + "num_tokens": 8127837.0, + "step": 1123 + }, + { + "epoch": 1.145769622833843, + "grad_norm": 5.6886305809021, + "learning_rate": 1e-06, + "loss": 0.3154, + "mean_token_accuracy": 0.8753517866134644, + "num_tokens": 8134804.0, + "step": 1124 + }, + { + "epoch": 1.146788990825688, + "grad_norm": 4.188750267028809, + "learning_rate": 1e-06, + "loss": 0.2681, + "mean_token_accuracy": 0.8964486718177795, + "num_tokens": 8143553.0, + "step": 1125 + }, + { + "epoch": 1.1478083588175332, + "grad_norm": 5.278468608856201, + "learning_rate": 1e-06, + "loss": 0.3072, + "mean_token_accuracy": 0.8796412348747253, + "num_tokens": 8149905.0, + "step": 1126 + }, + { + "epoch": 1.1488277268093783, + "grad_norm": 5.61359977722168, + "learning_rate": 1e-06, + "loss": 0.3221, + "mean_token_accuracy": 0.8656721115112305, + "num_tokens": 8156555.0, + "step": 1127 + }, + { + "epoch": 1.1498470948012232, + "grad_norm": 5.133378982543945, + "learning_rate": 1e-06, + "loss": 0.3419, + "mean_token_accuracy": 0.8668083548545837, + "num_tokens": 8164382.0, + "step": 1128 + }, + { + "epoch": 1.1508664627930683, + "grad_norm": 5.508909702301025, + "learning_rate": 1e-06, + "loss": 0.3108, + "mean_token_accuracy": 0.881386399269104, + "num_tokens": 8172585.0, + "step": 1129 + }, + { + "epoch": 1.1518858307849134, + "grad_norm": 5.778375148773193, + "learning_rate": 1e-06, + "loss": 0.3283, + "mean_token_accuracy": 0.8770470023155212, + "num_tokens": 8180167.0, + "step": 1130 + }, + { + "epoch": 1.1529051987767585, + "grad_norm": 5.272749423980713, + "learning_rate": 1e-06, + "loss": 0.3277, + "mean_token_accuracy": 0.8669830560684204, + "num_tokens": 8187543.0, + "step": 1131 + }, + { + "epoch": 1.1539245667686036, + "grad_norm": 5.614512920379639, + "learning_rate": 1e-06, + "loss": 0.2904, + "mean_token_accuracy": 0.8867930173873901, + "num_tokens": 8194131.0, + "step": 1132 + }, + { + "epoch": 1.1549439347604484, + "grad_norm": 5.642962455749512, + "learning_rate": 1e-06, + "loss": 0.3187, + "mean_token_accuracy": 0.872779369354248, + "num_tokens": 8201855.0, + "step": 1133 + }, + { + "epoch": 1.1559633027522935, + "grad_norm": 5.701113700866699, + "learning_rate": 1e-06, + "loss": 0.3125, + "mean_token_accuracy": 0.8772579431533813, + "num_tokens": 8208488.0, + "step": 1134 + }, + { + "epoch": 1.1569826707441386, + "grad_norm": 6.327109336853027, + "learning_rate": 1e-06, + "loss": 0.3564, + "mean_token_accuracy": 0.8517950773239136, + "num_tokens": 8214647.0, + "step": 1135 + }, + { + "epoch": 1.1580020387359837, + "grad_norm": 6.254746913909912, + "learning_rate": 1e-06, + "loss": 0.3255, + "mean_token_accuracy": 0.866521954536438, + "num_tokens": 8221024.0, + "step": 1136 + }, + { + "epoch": 1.1590214067278288, + "grad_norm": 5.454446315765381, + "learning_rate": 1e-06, + "loss": 0.3214, + "mean_token_accuracy": 0.8736889958381653, + "num_tokens": 8228649.0, + "step": 1137 + }, + { + "epoch": 1.1600407747196737, + "grad_norm": 5.65540075302124, + "learning_rate": 1e-06, + "loss": 0.3366, + "mean_token_accuracy": 0.8667781949043274, + "num_tokens": 8236400.0, + "step": 1138 + }, + { + "epoch": 1.1610601427115188, + "grad_norm": 6.349628448486328, + "learning_rate": 1e-06, + "loss": 0.3186, + "mean_token_accuracy": 0.871401309967041, + "num_tokens": 8241979.0, + "step": 1139 + }, + { + "epoch": 1.162079510703364, + "grad_norm": 6.348138809204102, + "learning_rate": 1e-06, + "loss": 0.3614, + "mean_token_accuracy": 0.858551025390625, + "num_tokens": 8247990.0, + "step": 1140 + }, + { + "epoch": 1.163098878695209, + "grad_norm": 4.625319004058838, + "learning_rate": 1e-06, + "loss": 0.2854, + "mean_token_accuracy": 0.8897217512130737, + "num_tokens": 8256295.0, + "step": 1141 + }, + { + "epoch": 1.164118246687054, + "grad_norm": 5.95109224319458, + "learning_rate": 1e-06, + "loss": 0.3099, + "mean_token_accuracy": 0.8800825476646423, + "num_tokens": 8262467.0, + "step": 1142 + }, + { + "epoch": 1.165137614678899, + "grad_norm": 5.6551594734191895, + "learning_rate": 1e-06, + "loss": 0.2984, + "mean_token_accuracy": 0.8798377513885498, + "num_tokens": 8268976.0, + "step": 1143 + }, + { + "epoch": 1.166156982670744, + "grad_norm": 5.972011566162109, + "learning_rate": 1e-06, + "loss": 0.3205, + "mean_token_accuracy": 0.8775017857551575, + "num_tokens": 8275471.0, + "step": 1144 + }, + { + "epoch": 1.1671763506625892, + "grad_norm": 5.007076740264893, + "learning_rate": 1e-06, + "loss": 0.2611, + "mean_token_accuracy": 0.8960992097854614, + "num_tokens": 8283238.0, + "step": 1145 + }, + { + "epoch": 1.1681957186544343, + "grad_norm": 5.306046962738037, + "learning_rate": 1e-06, + "loss": 0.2899, + "mean_token_accuracy": 0.8866194486618042, + "num_tokens": 8290573.0, + "step": 1146 + }, + { + "epoch": 1.1692150866462794, + "grad_norm": 4.894188404083252, + "learning_rate": 1e-06, + "loss": 0.2863, + "mean_token_accuracy": 0.8853701949119568, + "num_tokens": 8299140.0, + "step": 1147 + }, + { + "epoch": 1.1702344546381243, + "grad_norm": 5.633877754211426, + "learning_rate": 1e-06, + "loss": 0.3045, + "mean_token_accuracy": 0.8776700496673584, + "num_tokens": 8305439.0, + "step": 1148 + }, + { + "epoch": 1.1712538226299694, + "grad_norm": 4.756198406219482, + "learning_rate": 1e-06, + "loss": 0.2745, + "mean_token_accuracy": 0.8916192054748535, + "num_tokens": 8313651.0, + "step": 1149 + }, + { + "epoch": 1.1722731906218145, + "grad_norm": 5.658676624298096, + "learning_rate": 1e-06, + "loss": 0.2951, + "mean_token_accuracy": 0.8787216544151306, + "num_tokens": 8320670.0, + "step": 1150 + }, + { + "epoch": 1.1732925586136596, + "grad_norm": 5.555269718170166, + "learning_rate": 1e-06, + "loss": 0.3175, + "mean_token_accuracy": 0.8766881227493286, + "num_tokens": 8327491.0, + "step": 1151 + }, + { + "epoch": 1.1743119266055047, + "grad_norm": 6.442179203033447, + "learning_rate": 1e-06, + "loss": 0.34, + "mean_token_accuracy": 0.8691614866256714, + "num_tokens": 8333841.0, + "step": 1152 + }, + { + "epoch": 1.1753312945973495, + "grad_norm": 5.415842056274414, + "learning_rate": 1e-06, + "loss": 0.3167, + "mean_token_accuracy": 0.8810131549835205, + "num_tokens": 8341746.0, + "step": 1153 + }, + { + "epoch": 1.1763506625891946, + "grad_norm": 4.95771598815918, + "learning_rate": 1e-06, + "loss": 0.2947, + "mean_token_accuracy": 0.8778965473175049, + "num_tokens": 8350578.0, + "step": 1154 + }, + { + "epoch": 1.1773700305810397, + "grad_norm": 4.803190231323242, + "learning_rate": 1e-06, + "loss": 0.2647, + "mean_token_accuracy": 0.8935414552688599, + "num_tokens": 8358637.0, + "step": 1155 + }, + { + "epoch": 1.1783893985728848, + "grad_norm": 4.774295330047607, + "learning_rate": 1e-06, + "loss": 0.2697, + "mean_token_accuracy": 0.8951888084411621, + "num_tokens": 8366563.0, + "step": 1156 + }, + { + "epoch": 1.17940876656473, + "grad_norm": 5.012153625488281, + "learning_rate": 1e-06, + "loss": 0.3352, + "mean_token_accuracy": 0.8655050992965698, + "num_tokens": 8374034.0, + "step": 1157 + }, + { + "epoch": 1.1804281345565748, + "grad_norm": 6.14100980758667, + "learning_rate": 1e-06, + "loss": 0.3163, + "mean_token_accuracy": 0.8774555921554565, + "num_tokens": 8380178.0, + "step": 1158 + }, + { + "epoch": 1.18144750254842, + "grad_norm": 4.962352752685547, + "learning_rate": 1e-06, + "loss": 0.3203, + "mean_token_accuracy": 0.8765950202941895, + "num_tokens": 8389127.0, + "step": 1159 + }, + { + "epoch": 1.182466870540265, + "grad_norm": 5.9745283126831055, + "learning_rate": 1e-06, + "loss": 0.3167, + "mean_token_accuracy": 0.8743522763252258, + "num_tokens": 8395593.0, + "step": 1160 + }, + { + "epoch": 1.18348623853211, + "grad_norm": 6.105009078979492, + "learning_rate": 1e-06, + "loss": 0.328, + "mean_token_accuracy": 0.8697215914726257, + "num_tokens": 8401835.0, + "step": 1161 + }, + { + "epoch": 1.1845056065239552, + "grad_norm": 5.687469005584717, + "learning_rate": 1e-06, + "loss": 0.3094, + "mean_token_accuracy": 0.8817797899246216, + "num_tokens": 8409358.0, + "step": 1162 + }, + { + "epoch": 1.1855249745158003, + "grad_norm": 6.291867733001709, + "learning_rate": 1e-06, + "loss": 0.3458, + "mean_token_accuracy": 0.8659952878952026, + "num_tokens": 8415553.0, + "step": 1163 + }, + { + "epoch": 1.1865443425076452, + "grad_norm": 5.132673740386963, + "learning_rate": 1e-06, + "loss": 0.3089, + "mean_token_accuracy": 0.8772186040878296, + "num_tokens": 8423042.0, + "step": 1164 + }, + { + "epoch": 1.1875637104994903, + "grad_norm": 6.257059574127197, + "learning_rate": 1e-06, + "loss": 0.2944, + "mean_token_accuracy": 0.8802026510238647, + "num_tokens": 8430298.0, + "step": 1165 + }, + { + "epoch": 1.1885830784913354, + "grad_norm": 6.1363091468811035, + "learning_rate": 1e-06, + "loss": 0.3296, + "mean_token_accuracy": 0.8644047379493713, + "num_tokens": 8436781.0, + "step": 1166 + }, + { + "epoch": 1.1896024464831805, + "grad_norm": 5.754739284515381, + "learning_rate": 1e-06, + "loss": 0.2897, + "mean_token_accuracy": 0.8896462321281433, + "num_tokens": 8443077.0, + "step": 1167 + }, + { + "epoch": 1.1906218144750256, + "grad_norm": 5.319113254547119, + "learning_rate": 1e-06, + "loss": 0.2877, + "mean_token_accuracy": 0.8802158832550049, + "num_tokens": 8449626.0, + "step": 1168 + }, + { + "epoch": 1.1916411824668705, + "grad_norm": 4.94247579574585, + "learning_rate": 1e-06, + "loss": 0.314, + "mean_token_accuracy": 0.8760676383972168, + "num_tokens": 8458080.0, + "step": 1169 + }, + { + "epoch": 1.1926605504587156, + "grad_norm": 5.450106143951416, + "learning_rate": 1e-06, + "loss": 0.3179, + "mean_token_accuracy": 0.88019859790802, + "num_tokens": 8464511.0, + "step": 1170 + }, + { + "epoch": 1.1936799184505607, + "grad_norm": 5.956898212432861, + "learning_rate": 1e-06, + "loss": 0.3558, + "mean_token_accuracy": 0.861567497253418, + "num_tokens": 8471855.0, + "step": 1171 + }, + { + "epoch": 1.1946992864424058, + "grad_norm": 5.944676399230957, + "learning_rate": 1e-06, + "loss": 0.3088, + "mean_token_accuracy": 0.8803492188453674, + "num_tokens": 8478562.0, + "step": 1172 + }, + { + "epoch": 1.1957186544342508, + "grad_norm": 5.655628204345703, + "learning_rate": 1e-06, + "loss": 0.3234, + "mean_token_accuracy": 0.8713240027427673, + "num_tokens": 8485369.0, + "step": 1173 + }, + { + "epoch": 1.1967380224260957, + "grad_norm": 4.990614891052246, + "learning_rate": 1e-06, + "loss": 0.2824, + "mean_token_accuracy": 0.8765249252319336, + "num_tokens": 8492536.0, + "step": 1174 + }, + { + "epoch": 1.1977573904179408, + "grad_norm": 7.136660575866699, + "learning_rate": 1e-06, + "loss": 0.3472, + "mean_token_accuracy": 0.8651530146598816, + "num_tokens": 8498143.0, + "step": 1175 + }, + { + "epoch": 1.198776758409786, + "grad_norm": 4.662042140960693, + "learning_rate": 1e-06, + "loss": 0.2916, + "mean_token_accuracy": 0.8848963379859924, + "num_tokens": 8506577.0, + "step": 1176 + }, + { + "epoch": 1.199796126401631, + "grad_norm": 4.95400333404541, + "learning_rate": 1e-06, + "loss": 0.2764, + "mean_token_accuracy": 0.8906883597373962, + "num_tokens": 8514383.0, + "step": 1177 + }, + { + "epoch": 1.2008154943934761, + "grad_norm": 6.426281452178955, + "learning_rate": 1e-06, + "loss": 0.3896, + "mean_token_accuracy": 0.848471999168396, + "num_tokens": 8520945.0, + "step": 1178 + }, + { + "epoch": 1.2018348623853212, + "grad_norm": 4.998628616333008, + "learning_rate": 1e-06, + "loss": 0.2876, + "mean_token_accuracy": 0.8926323652267456, + "num_tokens": 8528456.0, + "step": 1179 + }, + { + "epoch": 1.202854230377166, + "grad_norm": 5.401933670043945, + "learning_rate": 1e-06, + "loss": 0.341, + "mean_token_accuracy": 0.8716030120849609, + "num_tokens": 8535863.0, + "step": 1180 + }, + { + "epoch": 1.2038735983690112, + "grad_norm": 5.050750732421875, + "learning_rate": 1e-06, + "loss": 0.3127, + "mean_token_accuracy": 0.8792431950569153, + "num_tokens": 8542810.0, + "step": 1181 + }, + { + "epoch": 1.2048929663608563, + "grad_norm": 5.314053535461426, + "learning_rate": 1e-06, + "loss": 0.3095, + "mean_token_accuracy": 0.8780306577682495, + "num_tokens": 8549623.0, + "step": 1182 + }, + { + "epoch": 1.2059123343527014, + "grad_norm": 5.393852710723877, + "learning_rate": 1e-06, + "loss": 0.3047, + "mean_token_accuracy": 0.8749839067459106, + "num_tokens": 8557067.0, + "step": 1183 + }, + { + "epoch": 1.2069317023445465, + "grad_norm": 4.798880577087402, + "learning_rate": 1e-06, + "loss": 0.2601, + "mean_token_accuracy": 0.9023774266242981, + "num_tokens": 8564749.0, + "step": 1184 + }, + { + "epoch": 1.2079510703363914, + "grad_norm": 5.731462001800537, + "learning_rate": 1e-06, + "loss": 0.3129, + "mean_token_accuracy": 0.878106415271759, + "num_tokens": 8571161.0, + "step": 1185 + }, + { + "epoch": 1.2089704383282365, + "grad_norm": 5.801967620849609, + "learning_rate": 1e-06, + "loss": 0.3305, + "mean_token_accuracy": 0.8683038353919983, + "num_tokens": 8578803.0, + "step": 1186 + }, + { + "epoch": 1.2099898063200816, + "grad_norm": 5.103005409240723, + "learning_rate": 1e-06, + "loss": 0.2779, + "mean_token_accuracy": 0.8825472593307495, + "num_tokens": 8586963.0, + "step": 1187 + }, + { + "epoch": 1.2110091743119267, + "grad_norm": 4.8528923988342285, + "learning_rate": 1e-06, + "loss": 0.2998, + "mean_token_accuracy": 0.8640416860580444, + "num_tokens": 8595435.0, + "step": 1188 + }, + { + "epoch": 1.2120285423037718, + "grad_norm": 5.175604820251465, + "learning_rate": 1e-06, + "loss": 0.3089, + "mean_token_accuracy": 0.8774794340133667, + "num_tokens": 8603457.0, + "step": 1189 + }, + { + "epoch": 1.2130479102956166, + "grad_norm": 5.8867034912109375, + "learning_rate": 1e-06, + "loss": 0.3224, + "mean_token_accuracy": 0.870144248008728, + "num_tokens": 8609610.0, + "step": 1190 + }, + { + "epoch": 1.2140672782874617, + "grad_norm": 6.624192237854004, + "learning_rate": 1e-06, + "loss": 0.37, + "mean_token_accuracy": 0.8606284856796265, + "num_tokens": 8615902.0, + "step": 1191 + }, + { + "epoch": 1.2150866462793068, + "grad_norm": 5.0010294914245605, + "learning_rate": 1e-06, + "loss": 0.301, + "mean_token_accuracy": 0.888220489025116, + "num_tokens": 8623890.0, + "step": 1192 + }, + { + "epoch": 1.216106014271152, + "grad_norm": 5.529745578765869, + "learning_rate": 1e-06, + "loss": 0.3153, + "mean_token_accuracy": 0.8762643337249756, + "num_tokens": 8630778.0, + "step": 1193 + }, + { + "epoch": 1.217125382262997, + "grad_norm": 4.841268539428711, + "learning_rate": 1e-06, + "loss": 0.2796, + "mean_token_accuracy": 0.8948809504508972, + "num_tokens": 8638369.0, + "step": 1194 + }, + { + "epoch": 1.218144750254842, + "grad_norm": 5.815280437469482, + "learning_rate": 1e-06, + "loss": 0.3237, + "mean_token_accuracy": 0.8746606111526489, + "num_tokens": 8644873.0, + "step": 1195 + }, + { + "epoch": 1.219164118246687, + "grad_norm": 5.067146301269531, + "learning_rate": 1e-06, + "loss": 0.2865, + "mean_token_accuracy": 0.8874868750572205, + "num_tokens": 8652673.0, + "step": 1196 + }, + { + "epoch": 1.2201834862385321, + "grad_norm": 5.3455095291137695, + "learning_rate": 1e-06, + "loss": 0.3141, + "mean_token_accuracy": 0.8758618831634521, + "num_tokens": 8659112.0, + "step": 1197 + }, + { + "epoch": 1.2212028542303772, + "grad_norm": 4.665726661682129, + "learning_rate": 1e-06, + "loss": 0.3005, + "mean_token_accuracy": 0.8849743604660034, + "num_tokens": 8667766.0, + "step": 1198 + }, + { + "epoch": 1.2222222222222223, + "grad_norm": 5.973724365234375, + "learning_rate": 1e-06, + "loss": 0.3313, + "mean_token_accuracy": 0.8654865026473999, + "num_tokens": 8674657.0, + "step": 1199 + }, + { + "epoch": 1.2232415902140672, + "grad_norm": 4.779473304748535, + "learning_rate": 1e-06, + "loss": 0.2841, + "mean_token_accuracy": 0.8870408535003662, + "num_tokens": 8682949.0, + "step": 1200 + }, + { + "epoch": 1.2242609582059123, + "grad_norm": 5.80247163772583, + "learning_rate": 1e-06, + "loss": 0.3265, + "mean_token_accuracy": 0.8723058104515076, + "num_tokens": 8689655.0, + "step": 1201 + }, + { + "epoch": 1.2252803261977574, + "grad_norm": 6.187098979949951, + "learning_rate": 1e-06, + "loss": 0.3087, + "mean_token_accuracy": 0.8812763690948486, + "num_tokens": 8696667.0, + "step": 1202 + }, + { + "epoch": 1.2262996941896025, + "grad_norm": 5.954038143157959, + "learning_rate": 1e-06, + "loss": 0.3361, + "mean_token_accuracy": 0.8683714866638184, + "num_tokens": 8702758.0, + "step": 1203 + }, + { + "epoch": 1.2273190621814476, + "grad_norm": 5.875865936279297, + "learning_rate": 1e-06, + "loss": 0.32, + "mean_token_accuracy": 0.8730576634407043, + "num_tokens": 8708818.0, + "step": 1204 + }, + { + "epoch": 1.2283384301732925, + "grad_norm": 4.809959888458252, + "learning_rate": 1e-06, + "loss": 0.3088, + "mean_token_accuracy": 0.8795297145843506, + "num_tokens": 8716925.0, + "step": 1205 + }, + { + "epoch": 1.2293577981651376, + "grad_norm": 5.191193580627441, + "learning_rate": 1e-06, + "loss": 0.2946, + "mean_token_accuracy": 0.8857678771018982, + "num_tokens": 8725844.0, + "step": 1206 + }, + { + "epoch": 1.2303771661569827, + "grad_norm": 5.093484878540039, + "learning_rate": 1e-06, + "loss": 0.2677, + "mean_token_accuracy": 0.8933212161064148, + "num_tokens": 8733529.0, + "step": 1207 + }, + { + "epoch": 1.2313965341488278, + "grad_norm": 5.977301120758057, + "learning_rate": 1e-06, + "loss": 0.3415, + "mean_token_accuracy": 0.8661149740219116, + "num_tokens": 8740364.0, + "step": 1208 + }, + { + "epoch": 1.2324159021406729, + "grad_norm": 5.4199018478393555, + "learning_rate": 1e-06, + "loss": 0.3011, + "mean_token_accuracy": 0.8818114995956421, + "num_tokens": 8747228.0, + "step": 1209 + }, + { + "epoch": 1.2334352701325177, + "grad_norm": 6.070919990539551, + "learning_rate": 1e-06, + "loss": 0.3132, + "mean_token_accuracy": 0.8760190010070801, + "num_tokens": 8753560.0, + "step": 1210 + }, + { + "epoch": 1.2344546381243628, + "grad_norm": 5.440700054168701, + "learning_rate": 1e-06, + "loss": 0.3018, + "mean_token_accuracy": 0.8820136189460754, + "num_tokens": 8760862.0, + "step": 1211 + }, + { + "epoch": 1.235474006116208, + "grad_norm": 5.896306037902832, + "learning_rate": 1e-06, + "loss": 0.3111, + "mean_token_accuracy": 0.8782287836074829, + "num_tokens": 8767847.0, + "step": 1212 + }, + { + "epoch": 1.236493374108053, + "grad_norm": 6.024957656860352, + "learning_rate": 1e-06, + "loss": 0.2973, + "mean_token_accuracy": 0.8786388635635376, + "num_tokens": 8774344.0, + "step": 1213 + }, + { + "epoch": 1.2375127420998981, + "grad_norm": 4.788306713104248, + "learning_rate": 1e-06, + "loss": 0.2515, + "mean_token_accuracy": 0.8993980884552002, + "num_tokens": 8782737.0, + "step": 1214 + }, + { + "epoch": 1.238532110091743, + "grad_norm": 5.7658915519714355, + "learning_rate": 1e-06, + "loss": 0.2755, + "mean_token_accuracy": 0.8959761261940002, + "num_tokens": 8790200.0, + "step": 1215 + }, + { + "epoch": 1.2395514780835881, + "grad_norm": 6.5592241287231445, + "learning_rate": 1e-06, + "loss": 0.2857, + "mean_token_accuracy": 0.8888245820999146, + "num_tokens": 8796615.0, + "step": 1216 + }, + { + "epoch": 1.2405708460754332, + "grad_norm": 5.90798282623291, + "learning_rate": 1e-06, + "loss": 0.3226, + "mean_token_accuracy": 0.868989109992981, + "num_tokens": 8803212.0, + "step": 1217 + }, + { + "epoch": 1.2415902140672783, + "grad_norm": 5.382796764373779, + "learning_rate": 1e-06, + "loss": 0.2953, + "mean_token_accuracy": 0.881160318851471, + "num_tokens": 8810976.0, + "step": 1218 + }, + { + "epoch": 1.2426095820591234, + "grad_norm": 6.543401718139648, + "learning_rate": 1e-06, + "loss": 0.3306, + "mean_token_accuracy": 0.867435097694397, + "num_tokens": 8817042.0, + "step": 1219 + }, + { + "epoch": 1.2436289500509683, + "grad_norm": 5.146704196929932, + "learning_rate": 1e-06, + "loss": 0.2856, + "mean_token_accuracy": 0.8827039003372192, + "num_tokens": 8824669.0, + "step": 1220 + }, + { + "epoch": 1.2446483180428134, + "grad_norm": 4.659019470214844, + "learning_rate": 1e-06, + "loss": 0.278, + "mean_token_accuracy": 0.8894573450088501, + "num_tokens": 8833669.0, + "step": 1221 + }, + { + "epoch": 1.2456676860346585, + "grad_norm": 5.04242467880249, + "learning_rate": 1e-06, + "loss": 0.3023, + "mean_token_accuracy": 0.8785600066184998, + "num_tokens": 8841591.0, + "step": 1222 + }, + { + "epoch": 1.2466870540265036, + "grad_norm": 5.7574543952941895, + "learning_rate": 1e-06, + "loss": 0.3085, + "mean_token_accuracy": 0.8792108297348022, + "num_tokens": 8848356.0, + "step": 1223 + }, + { + "epoch": 1.2477064220183487, + "grad_norm": 5.316781044006348, + "learning_rate": 1e-06, + "loss": 0.2947, + "mean_token_accuracy": 0.8825445175170898, + "num_tokens": 8855694.0, + "step": 1224 + }, + { + "epoch": 1.2487257900101936, + "grad_norm": 5.759340286254883, + "learning_rate": 1e-06, + "loss": 0.2876, + "mean_token_accuracy": 0.887660562992096, + "num_tokens": 8862200.0, + "step": 1225 + }, + { + "epoch": 1.2497451580020387, + "grad_norm": 6.173131942749023, + "learning_rate": 1e-06, + "loss": 0.301, + "mean_token_accuracy": 0.878262996673584, + "num_tokens": 8867518.0, + "step": 1226 + }, + { + "epoch": 1.2507645259938838, + "grad_norm": 5.828710556030273, + "learning_rate": 1e-06, + "loss": 0.2848, + "mean_token_accuracy": 0.883159339427948, + "num_tokens": 8873451.0, + "step": 1227 + }, + { + "epoch": 1.2517838939857289, + "grad_norm": 4.777270317077637, + "learning_rate": 1e-06, + "loss": 0.2822, + "mean_token_accuracy": 0.8859632015228271, + "num_tokens": 8881178.0, + "step": 1228 + }, + { + "epoch": 1.252803261977574, + "grad_norm": 5.937575817108154, + "learning_rate": 1e-06, + "loss": 0.3259, + "mean_token_accuracy": 0.8715857863426208, + "num_tokens": 8887059.0, + "step": 1229 + }, + { + "epoch": 1.2538226299694188, + "grad_norm": 6.058931827545166, + "learning_rate": 1e-06, + "loss": 0.3529, + "mean_token_accuracy": 0.8679486513137817, + "num_tokens": 8893787.0, + "step": 1230 + }, + { + "epoch": 1.254841997961264, + "grad_norm": 4.984497547149658, + "learning_rate": 1e-06, + "loss": 0.3179, + "mean_token_accuracy": 0.8712621927261353, + "num_tokens": 8901989.0, + "step": 1231 + }, + { + "epoch": 1.255861365953109, + "grad_norm": 6.145595073699951, + "learning_rate": 1e-06, + "loss": 0.3139, + "mean_token_accuracy": 0.8763754367828369, + "num_tokens": 8908189.0, + "step": 1232 + }, + { + "epoch": 1.2568807339449541, + "grad_norm": 5.243453025817871, + "learning_rate": 1e-06, + "loss": 0.2811, + "mean_token_accuracy": 0.8928725719451904, + "num_tokens": 8915183.0, + "step": 1233 + }, + { + "epoch": 1.2579001019367992, + "grad_norm": 6.521552562713623, + "learning_rate": 1e-06, + "loss": 0.3273, + "mean_token_accuracy": 0.8737128973007202, + "num_tokens": 8921551.0, + "step": 1234 + }, + { + "epoch": 1.2589194699286441, + "grad_norm": 5.251443386077881, + "learning_rate": 1e-06, + "loss": 0.3275, + "mean_token_accuracy": 0.8745530843734741, + "num_tokens": 8928498.0, + "step": 1235 + }, + { + "epoch": 1.2599388379204892, + "grad_norm": 5.178695201873779, + "learning_rate": 1e-06, + "loss": 0.3041, + "mean_token_accuracy": 0.8808330297470093, + "num_tokens": 8936853.0, + "step": 1236 + }, + { + "epoch": 1.2609582059123343, + "grad_norm": 5.201408863067627, + "learning_rate": 1e-06, + "loss": 0.2965, + "mean_token_accuracy": 0.872555136680603, + "num_tokens": 8943627.0, + "step": 1237 + }, + { + "epoch": 1.2619775739041794, + "grad_norm": 5.470115661621094, + "learning_rate": 1e-06, + "loss": 0.3047, + "mean_token_accuracy": 0.8864067792892456, + "num_tokens": 8950747.0, + "step": 1238 + }, + { + "epoch": 1.2629969418960245, + "grad_norm": 5.735748767852783, + "learning_rate": 1e-06, + "loss": 0.3473, + "mean_token_accuracy": 0.8673442602157593, + "num_tokens": 8957940.0, + "step": 1239 + }, + { + "epoch": 1.2640163098878694, + "grad_norm": 5.794240951538086, + "learning_rate": 1e-06, + "loss": 0.3292, + "mean_token_accuracy": 0.8683996796607971, + "num_tokens": 8964493.0, + "step": 1240 + }, + { + "epoch": 1.2650356778797147, + "grad_norm": 5.636260509490967, + "learning_rate": 1e-06, + "loss": 0.3235, + "mean_token_accuracy": 0.8710212111473083, + "num_tokens": 8971302.0, + "step": 1241 + }, + { + "epoch": 1.2660550458715596, + "grad_norm": 5.683717727661133, + "learning_rate": 1e-06, + "loss": 0.2836, + "mean_token_accuracy": 0.8866576552391052, + "num_tokens": 8977702.0, + "step": 1242 + }, + { + "epoch": 1.2670744138634047, + "grad_norm": 5.808632850646973, + "learning_rate": 1e-06, + "loss": 0.339, + "mean_token_accuracy": 0.8708400726318359, + "num_tokens": 8984587.0, + "step": 1243 + }, + { + "epoch": 1.2680937818552498, + "grad_norm": 5.006231307983398, + "learning_rate": 1e-06, + "loss": 0.2917, + "mean_token_accuracy": 0.8851825594902039, + "num_tokens": 8992603.0, + "step": 1244 + }, + { + "epoch": 1.2691131498470947, + "grad_norm": 5.840571880340576, + "learning_rate": 1e-06, + "loss": 0.3609, + "mean_token_accuracy": 0.8558920621871948, + "num_tokens": 8999260.0, + "step": 1245 + }, + { + "epoch": 1.27013251783894, + "grad_norm": 5.147486209869385, + "learning_rate": 1e-06, + "loss": 0.3111, + "mean_token_accuracy": 0.8794264793395996, + "num_tokens": 9006612.0, + "step": 1246 + }, + { + "epoch": 1.2711518858307849, + "grad_norm": 4.937698841094971, + "learning_rate": 1e-06, + "loss": 0.2765, + "mean_token_accuracy": 0.8815735578536987, + "num_tokens": 9013879.0, + "step": 1247 + }, + { + "epoch": 1.27217125382263, + "grad_norm": 5.333900451660156, + "learning_rate": 1e-06, + "loss": 0.3139, + "mean_token_accuracy": 0.8753398060798645, + "num_tokens": 9021203.0, + "step": 1248 + }, + { + "epoch": 1.273190621814475, + "grad_norm": 6.120837688446045, + "learning_rate": 1e-06, + "loss": 0.3238, + "mean_token_accuracy": 0.8733769655227661, + "num_tokens": 9027987.0, + "step": 1249 + }, + { + "epoch": 1.2742099898063202, + "grad_norm": 4.855228900909424, + "learning_rate": 1e-06, + "loss": 0.2802, + "mean_token_accuracy": 0.8782005906105042, + "num_tokens": 9035881.0, + "step": 1250 + }, + { + "epoch": 1.2752293577981653, + "grad_norm": 16.833330154418945, + "learning_rate": 1e-06, + "loss": 0.2895, + "mean_token_accuracy": 0.8855157494544983, + "num_tokens": 9043082.0, + "step": 1251 + }, + { + "epoch": 1.2762487257900101, + "grad_norm": 6.356771469116211, + "learning_rate": 1e-06, + "loss": 0.3078, + "mean_token_accuracy": 0.8760757446289062, + "num_tokens": 9048491.0, + "step": 1252 + }, + { + "epoch": 1.2772680937818552, + "grad_norm": 5.748081684112549, + "learning_rate": 1e-06, + "loss": 0.3298, + "mean_token_accuracy": 0.8745597004890442, + "num_tokens": 9054744.0, + "step": 1253 + }, + { + "epoch": 1.2782874617737003, + "grad_norm": 4.889772891998291, + "learning_rate": 1e-06, + "loss": 0.2786, + "mean_token_accuracy": 0.8933988809585571, + "num_tokens": 9063183.0, + "step": 1254 + }, + { + "epoch": 1.2793068297655454, + "grad_norm": 7.2224836349487305, + "learning_rate": 1e-06, + "loss": 0.3233, + "mean_token_accuracy": 0.8714964389801025, + "num_tokens": 9069905.0, + "step": 1255 + }, + { + "epoch": 1.2803261977573905, + "grad_norm": 5.249338150024414, + "learning_rate": 1e-06, + "loss": 0.2973, + "mean_token_accuracy": 0.8820124864578247, + "num_tokens": 9077523.0, + "step": 1256 + }, + { + "epoch": 1.2813455657492354, + "grad_norm": 6.612705707550049, + "learning_rate": 1e-06, + "loss": 0.3467, + "mean_token_accuracy": 0.8666405081748962, + "num_tokens": 9083516.0, + "step": 1257 + }, + { + "epoch": 1.2823649337410805, + "grad_norm": 4.43801212310791, + "learning_rate": 1e-06, + "loss": 0.275, + "mean_token_accuracy": 0.8884467482566833, + "num_tokens": 9092126.0, + "step": 1258 + }, + { + "epoch": 1.2833843017329256, + "grad_norm": 4.833804130554199, + "learning_rate": 1e-06, + "loss": 0.2547, + "mean_token_accuracy": 0.8981896638870239, + "num_tokens": 9099221.0, + "step": 1259 + }, + { + "epoch": 1.2844036697247707, + "grad_norm": 5.10806131362915, + "learning_rate": 1e-06, + "loss": 0.2961, + "mean_token_accuracy": 0.8792136311531067, + "num_tokens": 9107054.0, + "step": 1260 + }, + { + "epoch": 1.2854230377166158, + "grad_norm": 6.118267059326172, + "learning_rate": 1e-06, + "loss": 0.3729, + "mean_token_accuracy": 0.8530951738357544, + "num_tokens": 9114052.0, + "step": 1261 + }, + { + "epoch": 1.2864424057084607, + "grad_norm": 5.065720081329346, + "learning_rate": 1e-06, + "loss": 0.3065, + "mean_token_accuracy": 0.8797698020935059, + "num_tokens": 9122535.0, + "step": 1262 + }, + { + "epoch": 1.2874617737003058, + "grad_norm": 4.903911113739014, + "learning_rate": 1e-06, + "loss": 0.2933, + "mean_token_accuracy": 0.8857835531234741, + "num_tokens": 9130092.0, + "step": 1263 + }, + { + "epoch": 1.2884811416921509, + "grad_norm": 5.052695274353027, + "learning_rate": 1e-06, + "loss": 0.3154, + "mean_token_accuracy": 0.8820996880531311, + "num_tokens": 9139723.0, + "step": 1264 + }, + { + "epoch": 1.289500509683996, + "grad_norm": 6.907862663269043, + "learning_rate": 1e-06, + "loss": 0.3428, + "mean_token_accuracy": 0.8703360557556152, + "num_tokens": 9147134.0, + "step": 1265 + }, + { + "epoch": 1.290519877675841, + "grad_norm": 4.930056095123291, + "learning_rate": 1e-06, + "loss": 0.2812, + "mean_token_accuracy": 0.8791615962982178, + "num_tokens": 9156183.0, + "step": 1266 + }, + { + "epoch": 1.291539245667686, + "grad_norm": 4.859375953674316, + "learning_rate": 1e-06, + "loss": 0.2837, + "mean_token_accuracy": 0.8827297687530518, + "num_tokens": 9164075.0, + "step": 1267 + }, + { + "epoch": 1.292558613659531, + "grad_norm": 5.23075008392334, + "learning_rate": 1e-06, + "loss": 0.3229, + "mean_token_accuracy": 0.8715176582336426, + "num_tokens": 9171381.0, + "step": 1268 + }, + { + "epoch": 1.2935779816513762, + "grad_norm": 5.373680591583252, + "learning_rate": 1e-06, + "loss": 0.3076, + "mean_token_accuracy": 0.883700966835022, + "num_tokens": 9179011.0, + "step": 1269 + }, + { + "epoch": 1.2945973496432213, + "grad_norm": 4.913280487060547, + "learning_rate": 1e-06, + "loss": 0.3023, + "mean_token_accuracy": 0.8738519549369812, + "num_tokens": 9187146.0, + "step": 1270 + }, + { + "epoch": 1.2956167176350664, + "grad_norm": 5.1965556144714355, + "learning_rate": 1e-06, + "loss": 0.3078, + "mean_token_accuracy": 0.875895619392395, + "num_tokens": 9193825.0, + "step": 1271 + }, + { + "epoch": 1.2966360856269112, + "grad_norm": 7.396392822265625, + "learning_rate": 1e-06, + "loss": 0.3221, + "mean_token_accuracy": 0.871028482913971, + "num_tokens": 9200740.0, + "step": 1272 + }, + { + "epoch": 1.2976554536187563, + "grad_norm": 4.47261381149292, + "learning_rate": 1e-06, + "loss": 0.2832, + "mean_token_accuracy": 0.890716552734375, + "num_tokens": 9209316.0, + "step": 1273 + }, + { + "epoch": 1.2986748216106014, + "grad_norm": 5.151971340179443, + "learning_rate": 1e-06, + "loss": 0.3315, + "mean_token_accuracy": 0.8750662803649902, + "num_tokens": 9217007.0, + "step": 1274 + }, + { + "epoch": 1.2996941896024465, + "grad_norm": 6.134887218475342, + "learning_rate": 1e-06, + "loss": 0.2914, + "mean_token_accuracy": 0.8902760744094849, + "num_tokens": 9223484.0, + "step": 1275 + }, + { + "epoch": 1.3007135575942916, + "grad_norm": 6.109288692474365, + "learning_rate": 1e-06, + "loss": 0.3122, + "mean_token_accuracy": 0.8749300837516785, + "num_tokens": 9230635.0, + "step": 1276 + }, + { + "epoch": 1.3017329255861365, + "grad_norm": 6.126446723937988, + "learning_rate": 1e-06, + "loss": 0.3212, + "mean_token_accuracy": 0.8800523281097412, + "num_tokens": 9236937.0, + "step": 1277 + }, + { + "epoch": 1.3027522935779816, + "grad_norm": 5.202679634094238, + "learning_rate": 1e-06, + "loss": 0.2844, + "mean_token_accuracy": 0.8898653984069824, + "num_tokens": 9245029.0, + "step": 1278 + }, + { + "epoch": 1.3037716615698267, + "grad_norm": 5.710994720458984, + "learning_rate": 1e-06, + "loss": 0.3132, + "mean_token_accuracy": 0.8733188509941101, + "num_tokens": 9252131.0, + "step": 1279 + }, + { + "epoch": 1.3047910295616718, + "grad_norm": 5.697324752807617, + "learning_rate": 1e-06, + "loss": 0.2884, + "mean_token_accuracy": 0.8842966556549072, + "num_tokens": 9258933.0, + "step": 1280 + }, + { + "epoch": 1.305810397553517, + "grad_norm": 5.5946784019470215, + "learning_rate": 1e-06, + "loss": 0.3024, + "mean_token_accuracy": 0.8810803890228271, + "num_tokens": 9266081.0, + "step": 1281 + }, + { + "epoch": 1.3068297655453618, + "grad_norm": 5.746272563934326, + "learning_rate": 1e-06, + "loss": 0.3051, + "mean_token_accuracy": 0.8764121532440186, + "num_tokens": 9272990.0, + "step": 1282 + }, + { + "epoch": 1.3078491335372069, + "grad_norm": 6.186853408813477, + "learning_rate": 1e-06, + "loss": 0.3162, + "mean_token_accuracy": 0.8778024911880493, + "num_tokens": 9279242.0, + "step": 1283 + }, + { + "epoch": 1.308868501529052, + "grad_norm": 6.057239532470703, + "learning_rate": 1e-06, + "loss": 0.3424, + "mean_token_accuracy": 0.8697425723075867, + "num_tokens": 9286097.0, + "step": 1284 + }, + { + "epoch": 1.309887869520897, + "grad_norm": 5.407607555389404, + "learning_rate": 1e-06, + "loss": 0.3096, + "mean_token_accuracy": 0.8711328506469727, + "num_tokens": 9292898.0, + "step": 1285 + }, + { + "epoch": 1.3109072375127422, + "grad_norm": 6.252201557159424, + "learning_rate": 1e-06, + "loss": 0.3013, + "mean_token_accuracy": 0.8827858567237854, + "num_tokens": 9298854.0, + "step": 1286 + }, + { + "epoch": 1.311926605504587, + "grad_norm": 5.371987819671631, + "learning_rate": 1e-06, + "loss": 0.2989, + "mean_token_accuracy": 0.889165997505188, + "num_tokens": 9305784.0, + "step": 1287 + }, + { + "epoch": 1.3129459734964322, + "grad_norm": 6.198319911956787, + "learning_rate": 1e-06, + "loss": 0.3491, + "mean_token_accuracy": 0.8664237260818481, + "num_tokens": 9312779.0, + "step": 1288 + }, + { + "epoch": 1.3139653414882773, + "grad_norm": 6.0005927085876465, + "learning_rate": 1e-06, + "loss": 0.3116, + "mean_token_accuracy": 0.8747009038925171, + "num_tokens": 9319015.0, + "step": 1289 + }, + { + "epoch": 1.3149847094801224, + "grad_norm": 5.182898044586182, + "learning_rate": 1e-06, + "loss": 0.3136, + "mean_token_accuracy": 0.8780444860458374, + "num_tokens": 9327128.0, + "step": 1290 + }, + { + "epoch": 1.3160040774719675, + "grad_norm": 4.411076545715332, + "learning_rate": 1e-06, + "loss": 0.2608, + "mean_token_accuracy": 0.891155481338501, + "num_tokens": 9335929.0, + "step": 1291 + }, + { + "epoch": 1.3170234454638123, + "grad_norm": 4.566842079162598, + "learning_rate": 1e-06, + "loss": 0.2857, + "mean_token_accuracy": 0.8899908065795898, + "num_tokens": 9345038.0, + "step": 1292 + }, + { + "epoch": 1.3180428134556574, + "grad_norm": 5.385453224182129, + "learning_rate": 1e-06, + "loss": 0.3074, + "mean_token_accuracy": 0.8798048496246338, + "num_tokens": 9351731.0, + "step": 1293 + }, + { + "epoch": 1.3190621814475025, + "grad_norm": 4.363722324371338, + "learning_rate": 1e-06, + "loss": 0.2639, + "mean_token_accuracy": 0.8879799842834473, + "num_tokens": 9360709.0, + "step": 1294 + }, + { + "epoch": 1.3200815494393476, + "grad_norm": 6.098427772521973, + "learning_rate": 1e-06, + "loss": 0.324, + "mean_token_accuracy": 0.8739143013954163, + "num_tokens": 9367007.0, + "step": 1295 + }, + { + "epoch": 1.3211009174311927, + "grad_norm": 6.241596698760986, + "learning_rate": 1e-06, + "loss": 0.3064, + "mean_token_accuracy": 0.880509614944458, + "num_tokens": 9373373.0, + "step": 1296 + }, + { + "epoch": 1.3221202854230376, + "grad_norm": 4.855636119842529, + "learning_rate": 1e-06, + "loss": 0.2978, + "mean_token_accuracy": 0.876271665096283, + "num_tokens": 9381806.0, + "step": 1297 + }, + { + "epoch": 1.3231396534148827, + "grad_norm": 6.026391983032227, + "learning_rate": 1e-06, + "loss": 0.3344, + "mean_token_accuracy": 0.8747654557228088, + "num_tokens": 9388822.0, + "step": 1298 + }, + { + "epoch": 1.3241590214067278, + "grad_norm": 6.0617547035217285, + "learning_rate": 1e-06, + "loss": 0.3311, + "mean_token_accuracy": 0.8677884340286255, + "num_tokens": 9395635.0, + "step": 1299 + }, + { + "epoch": 1.325178389398573, + "grad_norm": 4.7792582511901855, + "learning_rate": 1e-06, + "loss": 0.2612, + "mean_token_accuracy": 0.8961860537528992, + "num_tokens": 9403816.0, + "step": 1300 + }, + { + "epoch": 1.326197757390418, + "grad_norm": 5.2650465965271, + "learning_rate": 1e-06, + "loss": 0.3067, + "mean_token_accuracy": 0.8841221332550049, + "num_tokens": 9411660.0, + "step": 1301 + }, + { + "epoch": 1.3272171253822629, + "grad_norm": 5.99510383605957, + "learning_rate": 1e-06, + "loss": 0.3643, + "mean_token_accuracy": 0.856602132320404, + "num_tokens": 9418897.0, + "step": 1302 + }, + { + "epoch": 1.328236493374108, + "grad_norm": 5.026741027832031, + "learning_rate": 1e-06, + "loss": 0.3205, + "mean_token_accuracy": 0.8748887777328491, + "num_tokens": 9426257.0, + "step": 1303 + }, + { + "epoch": 1.329255861365953, + "grad_norm": 5.001603126525879, + "learning_rate": 1e-06, + "loss": 0.3007, + "mean_token_accuracy": 0.8860616683959961, + "num_tokens": 9434306.0, + "step": 1304 + }, + { + "epoch": 1.3302752293577982, + "grad_norm": 6.558139801025391, + "learning_rate": 1e-06, + "loss": 0.3432, + "mean_token_accuracy": 0.8649742603302002, + "num_tokens": 9440449.0, + "step": 1305 + }, + { + "epoch": 1.3312945973496433, + "grad_norm": 5.542664527893066, + "learning_rate": 1e-06, + "loss": 0.2939, + "mean_token_accuracy": 0.8804621696472168, + "num_tokens": 9448019.0, + "step": 1306 + }, + { + "epoch": 1.3323139653414882, + "grad_norm": 4.85421085357666, + "learning_rate": 1e-06, + "loss": 0.2952, + "mean_token_accuracy": 0.8831931948661804, + "num_tokens": 9455638.0, + "step": 1307 + }, + { + "epoch": 1.3333333333333333, + "grad_norm": 5.87947416305542, + "learning_rate": 1e-06, + "loss": 0.3395, + "mean_token_accuracy": 0.8658331036567688, + "num_tokens": 9462692.0, + "step": 1308 + }, + { + "epoch": 1.3343527013251784, + "grad_norm": 6.633615493774414, + "learning_rate": 1e-06, + "loss": 0.3679, + "mean_token_accuracy": 0.8500319123268127, + "num_tokens": 9468523.0, + "step": 1309 + }, + { + "epoch": 1.3353720693170235, + "grad_norm": 5.894384384155273, + "learning_rate": 1e-06, + "loss": 0.3105, + "mean_token_accuracy": 0.8762898445129395, + "num_tokens": 9475172.0, + "step": 1310 + }, + { + "epoch": 1.3363914373088686, + "grad_norm": 5.109742164611816, + "learning_rate": 1e-06, + "loss": 0.2722, + "mean_token_accuracy": 0.8935374021530151, + "num_tokens": 9482040.0, + "step": 1311 + }, + { + "epoch": 1.3374108053007134, + "grad_norm": 5.244778156280518, + "learning_rate": 1e-06, + "loss": 0.3166, + "mean_token_accuracy": 0.8723559379577637, + "num_tokens": 9490475.0, + "step": 1312 + }, + { + "epoch": 1.3384301732925585, + "grad_norm": 4.533447742462158, + "learning_rate": 1e-06, + "loss": 0.2775, + "mean_token_accuracy": 0.8926587104797363, + "num_tokens": 9499334.0, + "step": 1313 + }, + { + "epoch": 1.3394495412844036, + "grad_norm": 5.489295482635498, + "learning_rate": 1e-06, + "loss": 0.3087, + "mean_token_accuracy": 0.8845779895782471, + "num_tokens": 9506045.0, + "step": 1314 + }, + { + "epoch": 1.3404689092762487, + "grad_norm": 5.7072978019714355, + "learning_rate": 1e-06, + "loss": 0.3175, + "mean_token_accuracy": 0.8713788986206055, + "num_tokens": 9513092.0, + "step": 1315 + }, + { + "epoch": 1.3414882772680938, + "grad_norm": 5.72050666809082, + "learning_rate": 1e-06, + "loss": 0.3006, + "mean_token_accuracy": 0.8814094066619873, + "num_tokens": 9521047.0, + "step": 1316 + }, + { + "epoch": 1.3425076452599387, + "grad_norm": 5.380745887756348, + "learning_rate": 1e-06, + "loss": 0.2841, + "mean_token_accuracy": 0.8823679089546204, + "num_tokens": 9528367.0, + "step": 1317 + }, + { + "epoch": 1.343527013251784, + "grad_norm": 5.160865306854248, + "learning_rate": 1e-06, + "loss": 0.302, + "mean_token_accuracy": 0.8840668201446533, + "num_tokens": 9536273.0, + "step": 1318 + }, + { + "epoch": 1.344546381243629, + "grad_norm": 5.88925313949585, + "learning_rate": 1e-06, + "loss": 0.3343, + "mean_token_accuracy": 0.8711948394775391, + "num_tokens": 9544833.0, + "step": 1319 + }, + { + "epoch": 1.345565749235474, + "grad_norm": 6.0200653076171875, + "learning_rate": 1e-06, + "loss": 0.3381, + "mean_token_accuracy": 0.8671168088912964, + "num_tokens": 9550862.0, + "step": 1320 + }, + { + "epoch": 1.346585117227319, + "grad_norm": 5.49343204498291, + "learning_rate": 1e-06, + "loss": 0.2955, + "mean_token_accuracy": 0.8790711164474487, + "num_tokens": 9558532.0, + "step": 1321 + }, + { + "epoch": 1.347604485219164, + "grad_norm": 4.877411365509033, + "learning_rate": 1e-06, + "loss": 0.3026, + "mean_token_accuracy": 0.8796015977859497, + "num_tokens": 9566357.0, + "step": 1322 + }, + { + "epoch": 1.3486238532110093, + "grad_norm": 6.063729763031006, + "learning_rate": 1e-06, + "loss": 0.3569, + "mean_token_accuracy": 0.8566425442695618, + "num_tokens": 9573323.0, + "step": 1323 + }, + { + "epoch": 1.3496432212028542, + "grad_norm": 4.060133457183838, + "learning_rate": 1e-06, + "loss": 0.2444, + "mean_token_accuracy": 0.8990655541419983, + "num_tokens": 9582209.0, + "step": 1324 + }, + { + "epoch": 1.3506625891946993, + "grad_norm": 4.864992141723633, + "learning_rate": 1e-06, + "loss": 0.2734, + "mean_token_accuracy": 0.8931567668914795, + "num_tokens": 9590274.0, + "step": 1325 + }, + { + "epoch": 1.3516819571865444, + "grad_norm": 6.065687656402588, + "learning_rate": 1e-06, + "loss": 0.328, + "mean_token_accuracy": 0.8716456890106201, + "num_tokens": 9596125.0, + "step": 1326 + }, + { + "epoch": 1.3527013251783895, + "grad_norm": 5.789783000946045, + "learning_rate": 1e-06, + "loss": 0.2874, + "mean_token_accuracy": 0.8879227042198181, + "num_tokens": 9603061.0, + "step": 1327 + }, + { + "epoch": 1.3537206931702346, + "grad_norm": 6.044703006744385, + "learning_rate": 1e-06, + "loss": 0.3581, + "mean_token_accuracy": 0.8678088188171387, + "num_tokens": 9609815.0, + "step": 1328 + }, + { + "epoch": 1.3547400611620795, + "grad_norm": 4.793939113616943, + "learning_rate": 1e-06, + "loss": 0.3092, + "mean_token_accuracy": 0.8793067336082458, + "num_tokens": 9617321.0, + "step": 1329 + }, + { + "epoch": 1.3557594291539246, + "grad_norm": 5.022708892822266, + "learning_rate": 1e-06, + "loss": 0.3294, + "mean_token_accuracy": 0.8680764436721802, + "num_tokens": 9625635.0, + "step": 1330 + }, + { + "epoch": 1.3567787971457697, + "grad_norm": 5.94536018371582, + "learning_rate": 1e-06, + "loss": 0.3123, + "mean_token_accuracy": 0.8784748315811157, + "num_tokens": 9632575.0, + "step": 1331 + }, + { + "epoch": 1.3577981651376148, + "grad_norm": 6.033531665802002, + "learning_rate": 1e-06, + "loss": 0.3097, + "mean_token_accuracy": 0.8811981678009033, + "num_tokens": 9638870.0, + "step": 1332 + }, + { + "epoch": 1.3588175331294599, + "grad_norm": 4.917769432067871, + "learning_rate": 1e-06, + "loss": 0.2819, + "mean_token_accuracy": 0.8792707324028015, + "num_tokens": 9646742.0, + "step": 1333 + }, + { + "epoch": 1.3598369011213047, + "grad_norm": 5.049533367156982, + "learning_rate": 1e-06, + "loss": 0.3487, + "mean_token_accuracy": 0.8642973899841309, + "num_tokens": 9655023.0, + "step": 1334 + }, + { + "epoch": 1.3608562691131498, + "grad_norm": 5.1791815757751465, + "learning_rate": 1e-06, + "loss": 0.3167, + "mean_token_accuracy": 0.8766209483146667, + "num_tokens": 9662080.0, + "step": 1335 + }, + { + "epoch": 1.361875637104995, + "grad_norm": 6.302092552185059, + "learning_rate": 1e-06, + "loss": 0.3469, + "mean_token_accuracy": 0.8611171841621399, + "num_tokens": 9668542.0, + "step": 1336 + }, + { + "epoch": 1.36289500509684, + "grad_norm": 4.921839714050293, + "learning_rate": 1e-06, + "loss": 0.2767, + "mean_token_accuracy": 0.8875328302383423, + "num_tokens": 9676101.0, + "step": 1337 + }, + { + "epoch": 1.3639143730886851, + "grad_norm": 6.112334728240967, + "learning_rate": 1e-06, + "loss": 0.3146, + "mean_token_accuracy": 0.8810312747955322, + "num_tokens": 9682761.0, + "step": 1338 + }, + { + "epoch": 1.36493374108053, + "grad_norm": 5.6241631507873535, + "learning_rate": 1e-06, + "loss": 0.3338, + "mean_token_accuracy": 0.8681468963623047, + "num_tokens": 9689752.0, + "step": 1339 + }, + { + "epoch": 1.365953109072375, + "grad_norm": 6.546330451965332, + "learning_rate": 1e-06, + "loss": 0.3335, + "mean_token_accuracy": 0.8667035102844238, + "num_tokens": 9695574.0, + "step": 1340 + }, + { + "epoch": 1.3669724770642202, + "grad_norm": 4.911314487457275, + "learning_rate": 1e-06, + "loss": 0.2982, + "mean_token_accuracy": 0.8794729709625244, + "num_tokens": 9703651.0, + "step": 1341 + }, + { + "epoch": 1.3679918450560653, + "grad_norm": 4.603722095489502, + "learning_rate": 1e-06, + "loss": 0.28, + "mean_token_accuracy": 0.8913896679878235, + "num_tokens": 9711865.0, + "step": 1342 + }, + { + "epoch": 1.3690112130479104, + "grad_norm": 5.438601016998291, + "learning_rate": 1e-06, + "loss": 0.3177, + "mean_token_accuracy": 0.8728730082511902, + "num_tokens": 9720699.0, + "step": 1343 + }, + { + "epoch": 1.3700305810397553, + "grad_norm": 5.4989495277404785, + "learning_rate": 1e-06, + "loss": 0.2953, + "mean_token_accuracy": 0.8871047496795654, + "num_tokens": 9727439.0, + "step": 1344 + }, + { + "epoch": 1.3710499490316004, + "grad_norm": 6.561010360717773, + "learning_rate": 1e-06, + "loss": 0.3574, + "mean_token_accuracy": 0.8595250248908997, + "num_tokens": 9734024.0, + "step": 1345 + }, + { + "epoch": 1.3720693170234455, + "grad_norm": 5.524742126464844, + "learning_rate": 1e-06, + "loss": 0.3574, + "mean_token_accuracy": 0.864406943321228, + "num_tokens": 9741732.0, + "step": 1346 + }, + { + "epoch": 1.3730886850152906, + "grad_norm": 6.441030502319336, + "learning_rate": 1e-06, + "loss": 0.3518, + "mean_token_accuracy": 0.8566567897796631, + "num_tokens": 9748399.0, + "step": 1347 + }, + { + "epoch": 1.3741080530071357, + "grad_norm": 5.153365612030029, + "learning_rate": 1e-06, + "loss": 0.3022, + "mean_token_accuracy": 0.8819665908813477, + "num_tokens": 9755868.0, + "step": 1348 + }, + { + "epoch": 1.3751274209989806, + "grad_norm": 5.254662990570068, + "learning_rate": 1e-06, + "loss": 0.2968, + "mean_token_accuracy": 0.8866252303123474, + "num_tokens": 9763519.0, + "step": 1349 + }, + { + "epoch": 1.3761467889908257, + "grad_norm": 5.568261623382568, + "learning_rate": 1e-06, + "loss": 0.3586, + "mean_token_accuracy": 0.8648022413253784, + "num_tokens": 9770256.0, + "step": 1350 + }, + { + "epoch": 1.3771661569826708, + "grad_norm": 4.945352077484131, + "learning_rate": 1e-06, + "loss": 0.3113, + "mean_token_accuracy": 0.8750856518745422, + "num_tokens": 9778107.0, + "step": 1351 + }, + { + "epoch": 1.3781855249745159, + "grad_norm": 5.016182899475098, + "learning_rate": 1e-06, + "loss": 0.306, + "mean_token_accuracy": 0.8841968774795532, + "num_tokens": 9785654.0, + "step": 1352 + }, + { + "epoch": 1.379204892966361, + "grad_norm": 5.979974269866943, + "learning_rate": 1e-06, + "loss": 0.3656, + "mean_token_accuracy": 0.8610702753067017, + "num_tokens": 9793413.0, + "step": 1353 + }, + { + "epoch": 1.3802242609582058, + "grad_norm": 4.572784423828125, + "learning_rate": 1e-06, + "loss": 0.2741, + "mean_token_accuracy": 0.8931981921195984, + "num_tokens": 9801345.0, + "step": 1354 + }, + { + "epoch": 1.381243628950051, + "grad_norm": 6.463254928588867, + "learning_rate": 1e-06, + "loss": 0.341, + "mean_token_accuracy": 0.8649834990501404, + "num_tokens": 9807384.0, + "step": 1355 + }, + { + "epoch": 1.382262996941896, + "grad_norm": 5.898228645324707, + "learning_rate": 1e-06, + "loss": 0.3352, + "mean_token_accuracy": 0.8701440095901489, + "num_tokens": 9814366.0, + "step": 1356 + }, + { + "epoch": 1.3832823649337411, + "grad_norm": 5.706131935119629, + "learning_rate": 1e-06, + "loss": 0.2799, + "mean_token_accuracy": 0.8925530314445496, + "num_tokens": 9821999.0, + "step": 1357 + }, + { + "epoch": 1.3843017329255862, + "grad_norm": 4.50039005279541, + "learning_rate": 1e-06, + "loss": 0.2688, + "mean_token_accuracy": 0.8971922397613525, + "num_tokens": 9830727.0, + "step": 1358 + }, + { + "epoch": 1.385321100917431, + "grad_norm": 5.6410040855407715, + "learning_rate": 1e-06, + "loss": 0.3025, + "mean_token_accuracy": 0.8773293495178223, + "num_tokens": 9837383.0, + "step": 1359 + }, + { + "epoch": 1.3863404689092762, + "grad_norm": 5.199773788452148, + "learning_rate": 1e-06, + "loss": 0.2761, + "mean_token_accuracy": 0.8910870552062988, + "num_tokens": 9844475.0, + "step": 1360 + }, + { + "epoch": 1.3873598369011213, + "grad_norm": 6.03353214263916, + "learning_rate": 1e-06, + "loss": 0.3119, + "mean_token_accuracy": 0.8747889995574951, + "num_tokens": 9851403.0, + "step": 1361 + }, + { + "epoch": 1.3883792048929664, + "grad_norm": 6.146934509277344, + "learning_rate": 1e-06, + "loss": 0.3548, + "mean_token_accuracy": 0.8632664084434509, + "num_tokens": 9859008.0, + "step": 1362 + }, + { + "epoch": 1.3893985728848115, + "grad_norm": 6.389611721038818, + "learning_rate": 1e-06, + "loss": 0.2945, + "mean_token_accuracy": 0.8782606720924377, + "num_tokens": 9865292.0, + "step": 1363 + }, + { + "epoch": 1.3904179408766564, + "grad_norm": 6.417214870452881, + "learning_rate": 1e-06, + "loss": 0.3566, + "mean_token_accuracy": 0.8571792840957642, + "num_tokens": 9871724.0, + "step": 1364 + }, + { + "epoch": 1.3914373088685015, + "grad_norm": 6.504530429840088, + "learning_rate": 1e-06, + "loss": 0.3568, + "mean_token_accuracy": 0.8601244688034058, + "num_tokens": 9878268.0, + "step": 1365 + }, + { + "epoch": 1.3924566768603466, + "grad_norm": 5.5968828201293945, + "learning_rate": 1e-06, + "loss": 0.3138, + "mean_token_accuracy": 0.8762362003326416, + "num_tokens": 9885884.0, + "step": 1366 + }, + { + "epoch": 1.3934760448521917, + "grad_norm": 5.75049352645874, + "learning_rate": 1e-06, + "loss": 0.3097, + "mean_token_accuracy": 0.8720189929008484, + "num_tokens": 9892504.0, + "step": 1367 + }, + { + "epoch": 1.3944954128440368, + "grad_norm": 5.128068447113037, + "learning_rate": 1e-06, + "loss": 0.3137, + "mean_token_accuracy": 0.8727405071258545, + "num_tokens": 9900002.0, + "step": 1368 + }, + { + "epoch": 1.3955147808358817, + "grad_norm": 4.7834391593933105, + "learning_rate": 1e-06, + "loss": 0.2641, + "mean_token_accuracy": 0.8830705881118774, + "num_tokens": 9907814.0, + "step": 1369 + }, + { + "epoch": 1.3965341488277268, + "grad_norm": 6.413769245147705, + "learning_rate": 1e-06, + "loss": 0.3192, + "mean_token_accuracy": 0.8747800588607788, + "num_tokens": 9913916.0, + "step": 1370 + }, + { + "epoch": 1.3975535168195719, + "grad_norm": 5.5533952713012695, + "learning_rate": 1e-06, + "loss": 0.2647, + "mean_token_accuracy": 0.8970255851745605, + "num_tokens": 9920811.0, + "step": 1371 + }, + { + "epoch": 1.398572884811417, + "grad_norm": 5.091334819793701, + "learning_rate": 1e-06, + "loss": 0.2688, + "mean_token_accuracy": 0.8885917663574219, + "num_tokens": 9928465.0, + "step": 1372 + }, + { + "epoch": 1.399592252803262, + "grad_norm": 6.220073699951172, + "learning_rate": 1e-06, + "loss": 0.3108, + "mean_token_accuracy": 0.8786689043045044, + "num_tokens": 9934857.0, + "step": 1373 + }, + { + "epoch": 1.400611620795107, + "grad_norm": 5.236258029937744, + "learning_rate": 1e-06, + "loss": 0.3094, + "mean_token_accuracy": 0.8798720240592957, + "num_tokens": 9944346.0, + "step": 1374 + }, + { + "epoch": 1.401630988786952, + "grad_norm": 5.496508598327637, + "learning_rate": 1e-06, + "loss": 0.3041, + "mean_token_accuracy": 0.8798038959503174, + "num_tokens": 9951379.0, + "step": 1375 + }, + { + "epoch": 1.4026503567787971, + "grad_norm": 5.121621608734131, + "learning_rate": 1e-06, + "loss": 0.2749, + "mean_token_accuracy": 0.8894817233085632, + "num_tokens": 9958359.0, + "step": 1376 + }, + { + "epoch": 1.4036697247706422, + "grad_norm": 6.177034378051758, + "learning_rate": 1e-06, + "loss": 0.3312, + "mean_token_accuracy": 0.8744955062866211, + "num_tokens": 9964588.0, + "step": 1377 + }, + { + "epoch": 1.4046890927624873, + "grad_norm": 5.4300384521484375, + "learning_rate": 1e-06, + "loss": 0.3125, + "mean_token_accuracy": 0.8739977478981018, + "num_tokens": 9971952.0, + "step": 1378 + }, + { + "epoch": 1.4057084607543322, + "grad_norm": 6.265637397766113, + "learning_rate": 1e-06, + "loss": 0.3346, + "mean_token_accuracy": 0.8722543716430664, + "num_tokens": 9978817.0, + "step": 1379 + }, + { + "epoch": 1.4067278287461773, + "grad_norm": 5.901264190673828, + "learning_rate": 1e-06, + "loss": 0.3132, + "mean_token_accuracy": 0.8800017833709717, + "num_tokens": 9987177.0, + "step": 1380 + }, + { + "epoch": 1.4077471967380224, + "grad_norm": 4.650755882263184, + "learning_rate": 1e-06, + "loss": 0.26, + "mean_token_accuracy": 0.8966093063354492, + "num_tokens": 9996219.0, + "step": 1381 + }, + { + "epoch": 1.4087665647298675, + "grad_norm": 5.538363933563232, + "learning_rate": 1e-06, + "loss": 0.2798, + "mean_token_accuracy": 0.8858299851417542, + "num_tokens": 10002735.0, + "step": 1382 + }, + { + "epoch": 1.4097859327217126, + "grad_norm": 5.061850547790527, + "learning_rate": 1e-06, + "loss": 0.2785, + "mean_token_accuracy": 0.8921376466751099, + "num_tokens": 10010710.0, + "step": 1383 + }, + { + "epoch": 1.4108053007135575, + "grad_norm": 5.615255832672119, + "learning_rate": 1e-06, + "loss": 0.3262, + "mean_token_accuracy": 0.8737493753433228, + "num_tokens": 10017549.0, + "step": 1384 + }, + { + "epoch": 1.4118246687054026, + "grad_norm": 5.257705211639404, + "learning_rate": 1e-06, + "loss": 0.3263, + "mean_token_accuracy": 0.8692662715911865, + "num_tokens": 10024897.0, + "step": 1385 + }, + { + "epoch": 1.4128440366972477, + "grad_norm": 5.553685665130615, + "learning_rate": 1e-06, + "loss": 0.326, + "mean_token_accuracy": 0.8713369369506836, + "num_tokens": 10032795.0, + "step": 1386 + }, + { + "epoch": 1.4138634046890928, + "grad_norm": 6.155224800109863, + "learning_rate": 1e-06, + "loss": 0.3721, + "mean_token_accuracy": 0.850556492805481, + "num_tokens": 10039236.0, + "step": 1387 + }, + { + "epoch": 1.4148827726809379, + "grad_norm": 5.0745625495910645, + "learning_rate": 1e-06, + "loss": 0.3151, + "mean_token_accuracy": 0.8714667558670044, + "num_tokens": 10046911.0, + "step": 1388 + }, + { + "epoch": 1.4159021406727827, + "grad_norm": 5.773280143737793, + "learning_rate": 1e-06, + "loss": 0.2983, + "mean_token_accuracy": 0.8847931623458862, + "num_tokens": 10054416.0, + "step": 1389 + }, + { + "epoch": 1.4169215086646278, + "grad_norm": 5.4401116371154785, + "learning_rate": 1e-06, + "loss": 0.3242, + "mean_token_accuracy": 0.8663142919540405, + "num_tokens": 10061353.0, + "step": 1390 + }, + { + "epoch": 1.417940876656473, + "grad_norm": 4.084571361541748, + "learning_rate": 1e-06, + "loss": 0.2448, + "mean_token_accuracy": 0.9057000875473022, + "num_tokens": 10070718.0, + "step": 1391 + }, + { + "epoch": 1.418960244648318, + "grad_norm": 5.479786396026611, + "learning_rate": 1e-06, + "loss": 0.3282, + "mean_token_accuracy": 0.8766346573829651, + "num_tokens": 10078504.0, + "step": 1392 + }, + { + "epoch": 1.4199796126401631, + "grad_norm": 6.595552444458008, + "learning_rate": 1e-06, + "loss": 0.3634, + "mean_token_accuracy": 0.8546597957611084, + "num_tokens": 10084564.0, + "step": 1393 + }, + { + "epoch": 1.420998980632008, + "grad_norm": 5.781376838684082, + "learning_rate": 1e-06, + "loss": 0.3268, + "mean_token_accuracy": 0.87879878282547, + "num_tokens": 10091673.0, + "step": 1394 + }, + { + "epoch": 1.4220183486238533, + "grad_norm": 6.455414295196533, + "learning_rate": 1e-06, + "loss": 0.3312, + "mean_token_accuracy": 0.8719620704650879, + "num_tokens": 10097735.0, + "step": 1395 + }, + { + "epoch": 1.4230377166156982, + "grad_norm": 5.796458721160889, + "learning_rate": 1e-06, + "loss": 0.3139, + "mean_token_accuracy": 0.877031683921814, + "num_tokens": 10105080.0, + "step": 1396 + }, + { + "epoch": 1.4240570846075433, + "grad_norm": 6.411151885986328, + "learning_rate": 1e-06, + "loss": 0.3413, + "mean_token_accuracy": 0.8713866472244263, + "num_tokens": 10110996.0, + "step": 1397 + }, + { + "epoch": 1.4250764525993884, + "grad_norm": 5.326949119567871, + "learning_rate": 1e-06, + "loss": 0.303, + "mean_token_accuracy": 0.8855444192886353, + "num_tokens": 10118583.0, + "step": 1398 + }, + { + "epoch": 1.4260958205912333, + "grad_norm": 4.630245208740234, + "learning_rate": 1e-06, + "loss": 0.2636, + "mean_token_accuracy": 0.8950007557868958, + "num_tokens": 10126867.0, + "step": 1399 + }, + { + "epoch": 1.4271151885830786, + "grad_norm": 5.294196128845215, + "learning_rate": 1e-06, + "loss": 0.3223, + "mean_token_accuracy": 0.8697159886360168, + "num_tokens": 10134786.0, + "step": 1400 + }, + { + "epoch": 1.4281345565749235, + "grad_norm": 4.45103645324707, + "learning_rate": 1e-06, + "loss": 0.284, + "mean_token_accuracy": 0.8904306888580322, + "num_tokens": 10143432.0, + "step": 1401 + }, + { + "epoch": 1.4291539245667686, + "grad_norm": 5.539140701293945, + "learning_rate": 1e-06, + "loss": 0.3258, + "mean_token_accuracy": 0.8748892545700073, + "num_tokens": 10149904.0, + "step": 1402 + }, + { + "epoch": 1.4301732925586137, + "grad_norm": 5.199077129364014, + "learning_rate": 1e-06, + "loss": 0.3267, + "mean_token_accuracy": 0.869633674621582, + "num_tokens": 10157127.0, + "step": 1403 + }, + { + "epoch": 1.4311926605504588, + "grad_norm": 5.123672962188721, + "learning_rate": 1e-06, + "loss": 0.2715, + "mean_token_accuracy": 0.8969149589538574, + "num_tokens": 10164184.0, + "step": 1404 + }, + { + "epoch": 1.432212028542304, + "grad_norm": 6.325544357299805, + "learning_rate": 1e-06, + "loss": 0.3173, + "mean_token_accuracy": 0.8781059980392456, + "num_tokens": 10169968.0, + "step": 1405 + }, + { + "epoch": 1.4332313965341488, + "grad_norm": 4.997438430786133, + "learning_rate": 1e-06, + "loss": 0.2833, + "mean_token_accuracy": 0.8824480772018433, + "num_tokens": 10177879.0, + "step": 1406 + }, + { + "epoch": 1.4342507645259939, + "grad_norm": 6.785586833953857, + "learning_rate": 1e-06, + "loss": 0.3389, + "mean_token_accuracy": 0.8698758482933044, + "num_tokens": 10185774.0, + "step": 1407 + }, + { + "epoch": 1.435270132517839, + "grad_norm": 5.316773414611816, + "learning_rate": 1e-06, + "loss": 0.2862, + "mean_token_accuracy": 0.8852212429046631, + "num_tokens": 10192780.0, + "step": 1408 + }, + { + "epoch": 1.436289500509684, + "grad_norm": 6.285282611846924, + "learning_rate": 1e-06, + "loss": 0.3266, + "mean_token_accuracy": 0.8709341287612915, + "num_tokens": 10199172.0, + "step": 1409 + }, + { + "epoch": 1.4373088685015292, + "grad_norm": 5.852154731750488, + "learning_rate": 1e-06, + "loss": 0.3377, + "mean_token_accuracy": 0.8716156482696533, + "num_tokens": 10207389.0, + "step": 1410 + }, + { + "epoch": 1.438328236493374, + "grad_norm": 6.212357521057129, + "learning_rate": 1e-06, + "loss": 0.3134, + "mean_token_accuracy": 0.8772020936012268, + "num_tokens": 10213696.0, + "step": 1411 + }, + { + "epoch": 1.4393476044852191, + "grad_norm": 5.483117580413818, + "learning_rate": 1e-06, + "loss": 0.3345, + "mean_token_accuracy": 0.8687236309051514, + "num_tokens": 10220771.0, + "step": 1412 + }, + { + "epoch": 1.4403669724770642, + "grad_norm": 5.505747318267822, + "learning_rate": 1e-06, + "loss": 0.3117, + "mean_token_accuracy": 0.8743664622306824, + "num_tokens": 10228386.0, + "step": 1413 + }, + { + "epoch": 1.4413863404689093, + "grad_norm": 6.010931968688965, + "learning_rate": 1e-06, + "loss": 0.3351, + "mean_token_accuracy": 0.8708958625793457, + "num_tokens": 10234786.0, + "step": 1414 + }, + { + "epoch": 1.4424057084607544, + "grad_norm": 4.583047389984131, + "learning_rate": 1e-06, + "loss": 0.285, + "mean_token_accuracy": 0.8940207958221436, + "num_tokens": 10244058.0, + "step": 1415 + }, + { + "epoch": 1.4434250764525993, + "grad_norm": 6.881775379180908, + "learning_rate": 1e-06, + "loss": 0.3641, + "mean_token_accuracy": 0.8590379953384399, + "num_tokens": 10250171.0, + "step": 1416 + }, + { + "epoch": 1.4444444444444444, + "grad_norm": 5.698225021362305, + "learning_rate": 1e-06, + "loss": 0.2848, + "mean_token_accuracy": 0.8766723871231079, + "num_tokens": 10257516.0, + "step": 1417 + }, + { + "epoch": 1.4454638124362895, + "grad_norm": 5.052091121673584, + "learning_rate": 1e-06, + "loss": 0.3252, + "mean_token_accuracy": 0.8723951578140259, + "num_tokens": 10265097.0, + "step": 1418 + }, + { + "epoch": 1.4464831804281346, + "grad_norm": 5.211941242218018, + "learning_rate": 1e-06, + "loss": 0.3183, + "mean_token_accuracy": 0.8768543004989624, + "num_tokens": 10272897.0, + "step": 1419 + }, + { + "epoch": 1.4475025484199797, + "grad_norm": 5.760153770446777, + "learning_rate": 1e-06, + "loss": 0.2898, + "mean_token_accuracy": 0.887344479560852, + "num_tokens": 10278797.0, + "step": 1420 + }, + { + "epoch": 1.4485219164118246, + "grad_norm": 5.362037658691406, + "learning_rate": 1e-06, + "loss": 0.3223, + "mean_token_accuracy": 0.868482232093811, + "num_tokens": 10285640.0, + "step": 1421 + }, + { + "epoch": 1.4495412844036697, + "grad_norm": 5.411773681640625, + "learning_rate": 1e-06, + "loss": 0.2907, + "mean_token_accuracy": 0.8785629272460938, + "num_tokens": 10292791.0, + "step": 1422 + }, + { + "epoch": 1.4505606523955148, + "grad_norm": 5.443609237670898, + "learning_rate": 1e-06, + "loss": 0.3359, + "mean_token_accuracy": 0.870162308216095, + "num_tokens": 10299971.0, + "step": 1423 + }, + { + "epoch": 1.45158002038736, + "grad_norm": 5.783640384674072, + "learning_rate": 1e-06, + "loss": 0.3522, + "mean_token_accuracy": 0.8604093790054321, + "num_tokens": 10307683.0, + "step": 1424 + }, + { + "epoch": 1.452599388379205, + "grad_norm": 4.652244567871094, + "learning_rate": 1e-06, + "loss": 0.3144, + "mean_token_accuracy": 0.8744456768035889, + "num_tokens": 10316495.0, + "step": 1425 + }, + { + "epoch": 1.4536187563710499, + "grad_norm": 5.338801383972168, + "learning_rate": 1e-06, + "loss": 0.3226, + "mean_token_accuracy": 0.8757918477058411, + "num_tokens": 10323847.0, + "step": 1426 + }, + { + "epoch": 1.454638124362895, + "grad_norm": 5.221811771392822, + "learning_rate": 1e-06, + "loss": 0.3082, + "mean_token_accuracy": 0.8727733492851257, + "num_tokens": 10331677.0, + "step": 1427 + }, + { + "epoch": 1.45565749235474, + "grad_norm": 5.047715663909912, + "learning_rate": 1e-06, + "loss": 0.3069, + "mean_token_accuracy": 0.8843958973884583, + "num_tokens": 10340876.0, + "step": 1428 + }, + { + "epoch": 1.4566768603465852, + "grad_norm": 5.320705413818359, + "learning_rate": 1e-06, + "loss": 0.3251, + "mean_token_accuracy": 0.8802404403686523, + "num_tokens": 10348778.0, + "step": 1429 + }, + { + "epoch": 1.4576962283384303, + "grad_norm": 6.204141139984131, + "learning_rate": 1e-06, + "loss": 0.3283, + "mean_token_accuracy": 0.8759833574295044, + "num_tokens": 10354356.0, + "step": 1430 + }, + { + "epoch": 1.4587155963302751, + "grad_norm": 4.833412170410156, + "learning_rate": 1e-06, + "loss": 0.2597, + "mean_token_accuracy": 0.8922300338745117, + "num_tokens": 10362232.0, + "step": 1431 + }, + { + "epoch": 1.4597349643221202, + "grad_norm": 6.14038610458374, + "learning_rate": 1e-06, + "loss": 0.3314, + "mean_token_accuracy": 0.8659457564353943, + "num_tokens": 10368323.0, + "step": 1432 + }, + { + "epoch": 1.4607543323139653, + "grad_norm": 6.86485481262207, + "learning_rate": 1e-06, + "loss": 0.3644, + "mean_token_accuracy": 0.8536801338195801, + "num_tokens": 10374486.0, + "step": 1433 + }, + { + "epoch": 1.4617737003058104, + "grad_norm": 6.280906677246094, + "learning_rate": 1e-06, + "loss": 0.3455, + "mean_token_accuracy": 0.8675920367240906, + "num_tokens": 10380377.0, + "step": 1434 + }, + { + "epoch": 1.4627930682976555, + "grad_norm": 4.845773220062256, + "learning_rate": 1e-06, + "loss": 0.3094, + "mean_token_accuracy": 0.8808777332305908, + "num_tokens": 10389158.0, + "step": 1435 + }, + { + "epoch": 1.4638124362895004, + "grad_norm": 6.093085289001465, + "learning_rate": 1e-06, + "loss": 0.3205, + "mean_token_accuracy": 0.8768734931945801, + "num_tokens": 10395515.0, + "step": 1436 + }, + { + "epoch": 1.4648318042813455, + "grad_norm": 5.048287868499756, + "learning_rate": 1e-06, + "loss": 0.2945, + "mean_token_accuracy": 0.8864775896072388, + "num_tokens": 10403259.0, + "step": 1437 + }, + { + "epoch": 1.4658511722731906, + "grad_norm": 5.117747783660889, + "learning_rate": 1e-06, + "loss": 0.313, + "mean_token_accuracy": 0.8808594942092896, + "num_tokens": 10411154.0, + "step": 1438 + }, + { + "epoch": 1.4668705402650357, + "grad_norm": 5.541556358337402, + "learning_rate": 1e-06, + "loss": 0.3151, + "mean_token_accuracy": 0.8768776655197144, + "num_tokens": 10418762.0, + "step": 1439 + }, + { + "epoch": 1.4678899082568808, + "grad_norm": 5.945231914520264, + "learning_rate": 1e-06, + "loss": 0.3358, + "mean_token_accuracy": 0.8645511269569397, + "num_tokens": 10424658.0, + "step": 1440 + }, + { + "epoch": 1.4689092762487257, + "grad_norm": 6.1465325355529785, + "learning_rate": 1e-06, + "loss": 0.3364, + "mean_token_accuracy": 0.8670410513877869, + "num_tokens": 10430817.0, + "step": 1441 + }, + { + "epoch": 1.4699286442405708, + "grad_norm": 5.852504253387451, + "learning_rate": 1e-06, + "loss": 0.3552, + "mean_token_accuracy": 0.8581855297088623, + "num_tokens": 10438086.0, + "step": 1442 + }, + { + "epoch": 1.470948012232416, + "grad_norm": 5.495230197906494, + "learning_rate": 1e-06, + "loss": 0.3109, + "mean_token_accuracy": 0.8742683529853821, + "num_tokens": 10444696.0, + "step": 1443 + }, + { + "epoch": 1.471967380224261, + "grad_norm": 4.766312599182129, + "learning_rate": 1e-06, + "loss": 0.271, + "mean_token_accuracy": 0.8953629732131958, + "num_tokens": 10452712.0, + "step": 1444 + }, + { + "epoch": 1.472986748216106, + "grad_norm": 5.715914726257324, + "learning_rate": 1e-06, + "loss": 0.3226, + "mean_token_accuracy": 0.8763505220413208, + "num_tokens": 10459425.0, + "step": 1445 + }, + { + "epoch": 1.474006116207951, + "grad_norm": 5.966634273529053, + "learning_rate": 1e-06, + "loss": 0.3273, + "mean_token_accuracy": 0.8699008226394653, + "num_tokens": 10466008.0, + "step": 1446 + }, + { + "epoch": 1.475025484199796, + "grad_norm": 6.030257701873779, + "learning_rate": 1e-06, + "loss": 0.2737, + "mean_token_accuracy": 0.8845248222351074, + "num_tokens": 10473184.0, + "step": 1447 + }, + { + "epoch": 1.4760448521916412, + "grad_norm": 6.008072853088379, + "learning_rate": 1e-06, + "loss": 0.3155, + "mean_token_accuracy": 0.8804813027381897, + "num_tokens": 10480713.0, + "step": 1448 + }, + { + "epoch": 1.4770642201834863, + "grad_norm": 5.072213172912598, + "learning_rate": 1e-06, + "loss": 0.2862, + "mean_token_accuracy": 0.8878883123397827, + "num_tokens": 10487543.0, + "step": 1449 + }, + { + "epoch": 1.4780835881753314, + "grad_norm": 5.575222015380859, + "learning_rate": 1e-06, + "loss": 0.2967, + "mean_token_accuracy": 0.8832118511199951, + "num_tokens": 10494702.0, + "step": 1450 + }, + { + "epoch": 1.4791029561671762, + "grad_norm": 6.518644332885742, + "learning_rate": 1e-06, + "loss": 0.3273, + "mean_token_accuracy": 0.86983323097229, + "num_tokens": 10500702.0, + "step": 1451 + }, + { + "epoch": 1.4801223241590213, + "grad_norm": 6.045334339141846, + "learning_rate": 1e-06, + "loss": 0.3451, + "mean_token_accuracy": 0.8612995743751526, + "num_tokens": 10506715.0, + "step": 1452 + }, + { + "epoch": 1.4811416921508664, + "grad_norm": 6.67121696472168, + "learning_rate": 1e-06, + "loss": 0.3317, + "mean_token_accuracy": 0.8636797666549683, + "num_tokens": 10512180.0, + "step": 1453 + }, + { + "epoch": 1.4821610601427115, + "grad_norm": 5.185829162597656, + "learning_rate": 1e-06, + "loss": 0.282, + "mean_token_accuracy": 0.8821680545806885, + "num_tokens": 10519406.0, + "step": 1454 + }, + { + "epoch": 1.4831804281345566, + "grad_norm": 5.244289875030518, + "learning_rate": 1e-06, + "loss": 0.2727, + "mean_token_accuracy": 0.88475501537323, + "num_tokens": 10526847.0, + "step": 1455 + }, + { + "epoch": 1.4841997961264015, + "grad_norm": 4.498580455780029, + "learning_rate": 1e-06, + "loss": 0.2671, + "mean_token_accuracy": 0.8925068378448486, + "num_tokens": 10534808.0, + "step": 1456 + }, + { + "epoch": 1.4852191641182466, + "grad_norm": 4.8494343757629395, + "learning_rate": 1e-06, + "loss": 0.315, + "mean_token_accuracy": 0.8729621171951294, + "num_tokens": 10543383.0, + "step": 1457 + }, + { + "epoch": 1.4862385321100917, + "grad_norm": 5.635125637054443, + "learning_rate": 1e-06, + "loss": 0.3178, + "mean_token_accuracy": 0.8748096823692322, + "num_tokens": 10549800.0, + "step": 1458 + }, + { + "epoch": 1.4872579001019368, + "grad_norm": 5.903260231018066, + "learning_rate": 1e-06, + "loss": 0.3213, + "mean_token_accuracy": 0.8695743083953857, + "num_tokens": 10556809.0, + "step": 1459 + }, + { + "epoch": 1.488277268093782, + "grad_norm": 5.3247809410095215, + "learning_rate": 1e-06, + "loss": 0.3187, + "mean_token_accuracy": 0.873119592666626, + "num_tokens": 10563720.0, + "step": 1460 + }, + { + "epoch": 1.4892966360856268, + "grad_norm": 6.042374610900879, + "learning_rate": 1e-06, + "loss": 0.3111, + "mean_token_accuracy": 0.8714637756347656, + "num_tokens": 10570380.0, + "step": 1461 + }, + { + "epoch": 1.490316004077472, + "grad_norm": 5.936408996582031, + "learning_rate": 1e-06, + "loss": 0.337, + "mean_token_accuracy": 0.8640352487564087, + "num_tokens": 10577103.0, + "step": 1462 + }, + { + "epoch": 1.491335372069317, + "grad_norm": 4.9831743240356445, + "learning_rate": 1e-06, + "loss": 0.2942, + "mean_token_accuracy": 0.8863672614097595, + "num_tokens": 10584928.0, + "step": 1463 + }, + { + "epoch": 1.492354740061162, + "grad_norm": 5.711718559265137, + "learning_rate": 1e-06, + "loss": 0.2954, + "mean_token_accuracy": 0.8816708326339722, + "num_tokens": 10591379.0, + "step": 1464 + }, + { + "epoch": 1.4933741080530072, + "grad_norm": 5.474244117736816, + "learning_rate": 1e-06, + "loss": 0.3279, + "mean_token_accuracy": 0.8724320530891418, + "num_tokens": 10598998.0, + "step": 1465 + }, + { + "epoch": 1.494393476044852, + "grad_norm": 6.191292762756348, + "learning_rate": 1e-06, + "loss": 0.3442, + "mean_token_accuracy": 0.8680183291435242, + "num_tokens": 10605023.0, + "step": 1466 + }, + { + "epoch": 1.4954128440366974, + "grad_norm": 5.439304828643799, + "learning_rate": 1e-06, + "loss": 0.3075, + "mean_token_accuracy": 0.8788709044456482, + "num_tokens": 10612306.0, + "step": 1467 + }, + { + "epoch": 1.4964322120285423, + "grad_norm": 5.360145092010498, + "learning_rate": 1e-06, + "loss": 0.317, + "mean_token_accuracy": 0.8773632049560547, + "num_tokens": 10620026.0, + "step": 1468 + }, + { + "epoch": 1.4974515800203874, + "grad_norm": 5.396060943603516, + "learning_rate": 1e-06, + "loss": 0.3026, + "mean_token_accuracy": 0.8843367695808411, + "num_tokens": 10627461.0, + "step": 1469 + }, + { + "epoch": 1.4984709480122325, + "grad_norm": 5.550556659698486, + "learning_rate": 1e-06, + "loss": 0.3087, + "mean_token_accuracy": 0.8800705671310425, + "num_tokens": 10634805.0, + "step": 1470 + }, + { + "epoch": 1.4994903160040773, + "grad_norm": 5.52756404876709, + "learning_rate": 1e-06, + "loss": 0.335, + "mean_token_accuracy": 0.8720515966415405, + "num_tokens": 10643119.0, + "step": 1471 + }, + { + "epoch": 1.5005096839959227, + "grad_norm": 6.386322498321533, + "learning_rate": 1e-06, + "loss": 0.3007, + "mean_token_accuracy": 0.8906170129776001, + "num_tokens": 10648943.0, + "step": 1472 + }, + { + "epoch": 1.5015290519877675, + "grad_norm": 6.179497718811035, + "learning_rate": 1e-06, + "loss": 0.3339, + "mean_token_accuracy": 0.8702336549758911, + "num_tokens": 10654837.0, + "step": 1473 + }, + { + "epoch": 1.5025484199796126, + "grad_norm": 5.657271862030029, + "learning_rate": 1e-06, + "loss": 0.2685, + "mean_token_accuracy": 0.8918006420135498, + "num_tokens": 10662028.0, + "step": 1474 + }, + { + "epoch": 1.5035677879714577, + "grad_norm": 5.940585136413574, + "learning_rate": 1e-06, + "loss": 0.3261, + "mean_token_accuracy": 0.8675059080123901, + "num_tokens": 10668545.0, + "step": 1475 + }, + { + "epoch": 1.5045871559633026, + "grad_norm": 5.752741813659668, + "learning_rate": 1e-06, + "loss": 0.3116, + "mean_token_accuracy": 0.8810972571372986, + "num_tokens": 10675602.0, + "step": 1476 + }, + { + "epoch": 1.505606523955148, + "grad_norm": 5.268730163574219, + "learning_rate": 1e-06, + "loss": 0.2661, + "mean_token_accuracy": 0.893947958946228, + "num_tokens": 10683186.0, + "step": 1477 + }, + { + "epoch": 1.5066258919469928, + "grad_norm": 5.619435787200928, + "learning_rate": 1e-06, + "loss": 0.3491, + "mean_token_accuracy": 0.8648844957351685, + "num_tokens": 10691427.0, + "step": 1478 + }, + { + "epoch": 1.507645259938838, + "grad_norm": 5.246896743774414, + "learning_rate": 1e-06, + "loss": 0.3354, + "mean_token_accuracy": 0.8681004047393799, + "num_tokens": 10699356.0, + "step": 1479 + }, + { + "epoch": 1.508664627930683, + "grad_norm": 5.273087501525879, + "learning_rate": 1e-06, + "loss": 0.3063, + "mean_token_accuracy": 0.8759640455245972, + "num_tokens": 10706570.0, + "step": 1480 + }, + { + "epoch": 1.5096839959225279, + "grad_norm": 5.202531337738037, + "learning_rate": 1e-06, + "loss": 0.2984, + "mean_token_accuracy": 0.8833551406860352, + "num_tokens": 10713731.0, + "step": 1481 + }, + { + "epoch": 1.5107033639143732, + "grad_norm": 5.1146063804626465, + "learning_rate": 1e-06, + "loss": 0.3188, + "mean_token_accuracy": 0.8796099424362183, + "num_tokens": 10722388.0, + "step": 1482 + }, + { + "epoch": 1.511722731906218, + "grad_norm": 5.505798816680908, + "learning_rate": 1e-06, + "loss": 0.3307, + "mean_token_accuracy": 0.8676062822341919, + "num_tokens": 10729788.0, + "step": 1483 + }, + { + "epoch": 1.5127420998980632, + "grad_norm": 5.962894916534424, + "learning_rate": 1e-06, + "loss": 0.3334, + "mean_token_accuracy": 0.8662010431289673, + "num_tokens": 10735888.0, + "step": 1484 + }, + { + "epoch": 1.5137614678899083, + "grad_norm": 6.091137409210205, + "learning_rate": 1e-06, + "loss": 0.3631, + "mean_token_accuracy": 0.8556003570556641, + "num_tokens": 10741969.0, + "step": 1485 + }, + { + "epoch": 1.5147808358817532, + "grad_norm": 5.25232458114624, + "learning_rate": 1e-06, + "loss": 0.3185, + "mean_token_accuracy": 0.8697114586830139, + "num_tokens": 10749574.0, + "step": 1486 + }, + { + "epoch": 1.5158002038735985, + "grad_norm": 4.114347457885742, + "learning_rate": 1e-06, + "loss": 0.2695, + "mean_token_accuracy": 0.8884599208831787, + "num_tokens": 10759668.0, + "step": 1487 + }, + { + "epoch": 1.5168195718654434, + "grad_norm": 5.790799140930176, + "learning_rate": 1e-06, + "loss": 0.2809, + "mean_token_accuracy": 0.891918420791626, + "num_tokens": 10765825.0, + "step": 1488 + }, + { + "epoch": 1.5178389398572885, + "grad_norm": 6.516838073730469, + "learning_rate": 1e-06, + "loss": 0.3274, + "mean_token_accuracy": 0.8740319013595581, + "num_tokens": 10772136.0, + "step": 1489 + }, + { + "epoch": 1.5188583078491336, + "grad_norm": 5.855703830718994, + "learning_rate": 1e-06, + "loss": 0.3148, + "mean_token_accuracy": 0.8771113157272339, + "num_tokens": 10779303.0, + "step": 1490 + }, + { + "epoch": 1.5198776758409784, + "grad_norm": 6.02843713760376, + "learning_rate": 1e-06, + "loss": 0.3316, + "mean_token_accuracy": 0.870262086391449, + "num_tokens": 10785156.0, + "step": 1491 + }, + { + "epoch": 1.5208970438328238, + "grad_norm": 7.142403602600098, + "learning_rate": 1e-06, + "loss": 0.3745, + "mean_token_accuracy": 0.8541925549507141, + "num_tokens": 10790863.0, + "step": 1492 + }, + { + "epoch": 1.5219164118246686, + "grad_norm": 5.865224361419678, + "learning_rate": 1e-06, + "loss": 0.3073, + "mean_token_accuracy": 0.8727608919143677, + "num_tokens": 10796830.0, + "step": 1493 + }, + { + "epoch": 1.5229357798165137, + "grad_norm": 5.133249282836914, + "learning_rate": 1e-06, + "loss": 0.299, + "mean_token_accuracy": 0.8769875764846802, + "num_tokens": 10804552.0, + "step": 1494 + }, + { + "epoch": 1.5239551478083588, + "grad_norm": 6.533811092376709, + "learning_rate": 1e-06, + "loss": 0.32, + "mean_token_accuracy": 0.8731846809387207, + "num_tokens": 10810398.0, + "step": 1495 + }, + { + "epoch": 1.5249745158002037, + "grad_norm": 6.476481914520264, + "learning_rate": 1e-06, + "loss": 0.2941, + "mean_token_accuracy": 0.8851321935653687, + "num_tokens": 10817722.0, + "step": 1496 + }, + { + "epoch": 1.525993883792049, + "grad_norm": 4.513082027435303, + "learning_rate": 1e-06, + "loss": 0.2698, + "mean_token_accuracy": 0.8900938630104065, + "num_tokens": 10825970.0, + "step": 1497 + }, + { + "epoch": 1.527013251783894, + "grad_norm": 6.015381813049316, + "learning_rate": 1e-06, + "loss": 0.3238, + "mean_token_accuracy": 0.875347912311554, + "num_tokens": 10832562.0, + "step": 1498 + }, + { + "epoch": 1.528032619775739, + "grad_norm": 6.958033561706543, + "learning_rate": 1e-06, + "loss": 0.3656, + "mean_token_accuracy": 0.8505918979644775, + "num_tokens": 10838209.0, + "step": 1499 + }, + { + "epoch": 1.529051987767584, + "grad_norm": 5.300173759460449, + "learning_rate": 1e-06, + "loss": 0.2997, + "mean_token_accuracy": 0.8818399906158447, + "num_tokens": 10846233.0, + "step": 1500 + }, + { + "epoch": 1.5300713557594292, + "grad_norm": 5.797749996185303, + "learning_rate": 1e-06, + "loss": 0.2926, + "mean_token_accuracy": 0.8802342414855957, + "num_tokens": 10852682.0, + "step": 1501 + }, + { + "epoch": 1.5310907237512743, + "grad_norm": 5.663289546966553, + "learning_rate": 1e-06, + "loss": 0.3232, + "mean_token_accuracy": 0.8728860020637512, + "num_tokens": 10859580.0, + "step": 1502 + }, + { + "epoch": 1.5321100917431192, + "grad_norm": 4.068572998046875, + "learning_rate": 1e-06, + "loss": 0.2606, + "mean_token_accuracy": 0.8924882411956787, + "num_tokens": 10869203.0, + "step": 1503 + }, + { + "epoch": 1.5331294597349643, + "grad_norm": 8.637890815734863, + "learning_rate": 1e-06, + "loss": 0.3308, + "mean_token_accuracy": 0.8720992803573608, + "num_tokens": 10876135.0, + "step": 1504 + }, + { + "epoch": 1.5341488277268094, + "grad_norm": 5.71559476852417, + "learning_rate": 1e-06, + "loss": 0.3383, + "mean_token_accuracy": 0.8672224283218384, + "num_tokens": 10882936.0, + "step": 1505 + }, + { + "epoch": 1.5351681957186545, + "grad_norm": 5.288494110107422, + "learning_rate": 1e-06, + "loss": 0.3101, + "mean_token_accuracy": 0.8821942806243896, + "num_tokens": 10890519.0, + "step": 1506 + }, + { + "epoch": 1.5361875637104996, + "grad_norm": 5.610746383666992, + "learning_rate": 1e-06, + "loss": 0.3421, + "mean_token_accuracy": 0.860659122467041, + "num_tokens": 10898139.0, + "step": 1507 + }, + { + "epoch": 1.5372069317023445, + "grad_norm": 6.0359320640563965, + "learning_rate": 1e-06, + "loss": 0.3355, + "mean_token_accuracy": 0.8690934181213379, + "num_tokens": 10904679.0, + "step": 1508 + }, + { + "epoch": 1.5382262996941896, + "grad_norm": 5.640620231628418, + "learning_rate": 1e-06, + "loss": 0.3111, + "mean_token_accuracy": 0.8787278532981873, + "num_tokens": 10911531.0, + "step": 1509 + }, + { + "epoch": 1.5392456676860347, + "grad_norm": 5.089776039123535, + "learning_rate": 1e-06, + "loss": 0.3212, + "mean_token_accuracy": 0.874971866607666, + "num_tokens": 10918978.0, + "step": 1510 + }, + { + "epoch": 1.5402650356778798, + "grad_norm": 5.368494987487793, + "learning_rate": 1e-06, + "loss": 0.3184, + "mean_token_accuracy": 0.8763478994369507, + "num_tokens": 10926844.0, + "step": 1511 + }, + { + "epoch": 1.5412844036697249, + "grad_norm": 5.884599208831787, + "learning_rate": 1e-06, + "loss": 0.2852, + "mean_token_accuracy": 0.8897175192832947, + "num_tokens": 10933478.0, + "step": 1512 + }, + { + "epoch": 1.5423037716615697, + "grad_norm": 7.0595784187316895, + "learning_rate": 1e-06, + "loss": 0.3866, + "mean_token_accuracy": 0.8540596961975098, + "num_tokens": 10939042.0, + "step": 1513 + }, + { + "epoch": 1.5433231396534148, + "grad_norm": 5.767219543457031, + "learning_rate": 1e-06, + "loss": 0.32, + "mean_token_accuracy": 0.8738173246383667, + "num_tokens": 10945243.0, + "step": 1514 + }, + { + "epoch": 1.54434250764526, + "grad_norm": 5.82436466217041, + "learning_rate": 1e-06, + "loss": 0.3405, + "mean_token_accuracy": 0.8605579137802124, + "num_tokens": 10951612.0, + "step": 1515 + }, + { + "epoch": 1.545361875637105, + "grad_norm": 5.540253162384033, + "learning_rate": 1e-06, + "loss": 0.329, + "mean_token_accuracy": 0.8672847151756287, + "num_tokens": 10958555.0, + "step": 1516 + }, + { + "epoch": 1.5463812436289501, + "grad_norm": 5.483942985534668, + "learning_rate": 1e-06, + "loss": 0.3334, + "mean_token_accuracy": 0.8655393123626709, + "num_tokens": 10965847.0, + "step": 1517 + }, + { + "epoch": 1.547400611620795, + "grad_norm": 5.820426940917969, + "learning_rate": 1e-06, + "loss": 0.3252, + "mean_token_accuracy": 0.8738278150558472, + "num_tokens": 10972531.0, + "step": 1518 + }, + { + "epoch": 1.5484199796126403, + "grad_norm": 5.546191692352295, + "learning_rate": 1e-06, + "loss": 0.2822, + "mean_token_accuracy": 0.8876298666000366, + "num_tokens": 10979338.0, + "step": 1519 + }, + { + "epoch": 1.5494393476044852, + "grad_norm": 5.39108419418335, + "learning_rate": 1e-06, + "loss": 0.3126, + "mean_token_accuracy": 0.8795590996742249, + "num_tokens": 10986367.0, + "step": 1520 + }, + { + "epoch": 1.5504587155963303, + "grad_norm": 5.293471813201904, + "learning_rate": 1e-06, + "loss": 0.3151, + "mean_token_accuracy": 0.8807576298713684, + "num_tokens": 10994283.0, + "step": 1521 + }, + { + "epoch": 1.5514780835881754, + "grad_norm": 5.949269771575928, + "learning_rate": 1e-06, + "loss": 0.3572, + "mean_token_accuracy": 0.8548540472984314, + "num_tokens": 11001154.0, + "step": 1522 + }, + { + "epoch": 1.5524974515800203, + "grad_norm": 5.902353763580322, + "learning_rate": 1e-06, + "loss": 0.3519, + "mean_token_accuracy": 0.8621294498443604, + "num_tokens": 11007781.0, + "step": 1523 + }, + { + "epoch": 1.5535168195718656, + "grad_norm": 5.682134628295898, + "learning_rate": 1e-06, + "loss": 0.315, + "mean_token_accuracy": 0.8816536664962769, + "num_tokens": 11015137.0, + "step": 1524 + }, + { + "epoch": 1.5545361875637105, + "grad_norm": 4.910098552703857, + "learning_rate": 1e-06, + "loss": 0.2774, + "mean_token_accuracy": 0.8912296891212463, + "num_tokens": 11022972.0, + "step": 1525 + }, + { + "epoch": 1.5555555555555556, + "grad_norm": 6.068448066711426, + "learning_rate": 1e-06, + "loss": 0.3159, + "mean_token_accuracy": 0.8756000399589539, + "num_tokens": 11029062.0, + "step": 1526 + }, + { + "epoch": 1.5565749235474007, + "grad_norm": 5.475059509277344, + "learning_rate": 1e-06, + "loss": 0.3234, + "mean_token_accuracy": 0.8692553639411926, + "num_tokens": 11035425.0, + "step": 1527 + }, + { + "epoch": 1.5575942915392456, + "grad_norm": 5.3091535568237305, + "learning_rate": 1e-06, + "loss": 0.3037, + "mean_token_accuracy": 0.888388991355896, + "num_tokens": 11043012.0, + "step": 1528 + }, + { + "epoch": 1.5586136595310909, + "grad_norm": 6.23546838760376, + "learning_rate": 1e-06, + "loss": 0.3154, + "mean_token_accuracy": 0.8742067813873291, + "num_tokens": 11049754.0, + "step": 1529 + }, + { + "epoch": 1.5596330275229358, + "grad_norm": 6.140744686126709, + "learning_rate": 1e-06, + "loss": 0.3384, + "mean_token_accuracy": 0.8720318078994751, + "num_tokens": 11056547.0, + "step": 1530 + }, + { + "epoch": 1.5606523955147809, + "grad_norm": 5.692973613739014, + "learning_rate": 1e-06, + "loss": 0.3442, + "mean_token_accuracy": 0.8611264228820801, + "num_tokens": 11064035.0, + "step": 1531 + }, + { + "epoch": 1.561671763506626, + "grad_norm": 5.775223255157471, + "learning_rate": 1e-06, + "loss": 0.342, + "mean_token_accuracy": 0.8701015710830688, + "num_tokens": 11071055.0, + "step": 1532 + }, + { + "epoch": 1.5626911314984708, + "grad_norm": 5.839698314666748, + "learning_rate": 1e-06, + "loss": 0.3481, + "mean_token_accuracy": 0.8572856783866882, + "num_tokens": 11077924.0, + "step": 1533 + }, + { + "epoch": 1.5637104994903162, + "grad_norm": 5.289140224456787, + "learning_rate": 1e-06, + "loss": 0.3338, + "mean_token_accuracy": 0.870862603187561, + "num_tokens": 11085374.0, + "step": 1534 + }, + { + "epoch": 1.564729867482161, + "grad_norm": 4.831606864929199, + "learning_rate": 1e-06, + "loss": 0.2924, + "mean_token_accuracy": 0.8883854150772095, + "num_tokens": 11093030.0, + "step": 1535 + }, + { + "epoch": 1.5657492354740061, + "grad_norm": 4.691396236419678, + "learning_rate": 1e-06, + "loss": 0.2513, + "mean_token_accuracy": 0.9037646055221558, + "num_tokens": 11100232.0, + "step": 1536 + }, + { + "epoch": 1.5667686034658512, + "grad_norm": 5.540989875793457, + "learning_rate": 1e-06, + "loss": 0.3405, + "mean_token_accuracy": 0.86326003074646, + "num_tokens": 11106940.0, + "step": 1537 + }, + { + "epoch": 1.567787971457696, + "grad_norm": 5.794976711273193, + "learning_rate": 1e-06, + "loss": 0.2945, + "mean_token_accuracy": 0.879813551902771, + "num_tokens": 11114405.0, + "step": 1538 + }, + { + "epoch": 1.5688073394495414, + "grad_norm": 5.989882946014404, + "learning_rate": 1e-06, + "loss": 0.3048, + "mean_token_accuracy": 0.8844531178474426, + "num_tokens": 11121382.0, + "step": 1539 + }, + { + "epoch": 1.5698267074413863, + "grad_norm": 5.554046154022217, + "learning_rate": 1e-06, + "loss": 0.355, + "mean_token_accuracy": 0.8611112236976624, + "num_tokens": 11129303.0, + "step": 1540 + }, + { + "epoch": 1.5708460754332314, + "grad_norm": 5.9325079917907715, + "learning_rate": 1e-06, + "loss": 0.3363, + "mean_token_accuracy": 0.8599929213523865, + "num_tokens": 11136297.0, + "step": 1541 + }, + { + "epoch": 1.5718654434250765, + "grad_norm": 6.693843364715576, + "learning_rate": 1e-06, + "loss": 0.3425, + "mean_token_accuracy": 0.8640676736831665, + "num_tokens": 11142095.0, + "step": 1542 + }, + { + "epoch": 1.5728848114169214, + "grad_norm": 5.159486770629883, + "learning_rate": 1e-06, + "loss": 0.2818, + "mean_token_accuracy": 0.8912928104400635, + "num_tokens": 11150349.0, + "step": 1543 + }, + { + "epoch": 1.5739041794087667, + "grad_norm": 6.92703104019165, + "learning_rate": 1e-06, + "loss": 0.3744, + "mean_token_accuracy": 0.8501841425895691, + "num_tokens": 11155942.0, + "step": 1544 + }, + { + "epoch": 1.5749235474006116, + "grad_norm": 6.002547740936279, + "learning_rate": 1e-06, + "loss": 0.2925, + "mean_token_accuracy": 0.8841749429702759, + "num_tokens": 11162798.0, + "step": 1545 + }, + { + "epoch": 1.5759429153924567, + "grad_norm": 5.155179977416992, + "learning_rate": 1e-06, + "loss": 0.2897, + "mean_token_accuracy": 0.8933208584785461, + "num_tokens": 11169950.0, + "step": 1546 + }, + { + "epoch": 1.5769622833843018, + "grad_norm": 5.27327823638916, + "learning_rate": 1e-06, + "loss": 0.3424, + "mean_token_accuracy": 0.8611549139022827, + "num_tokens": 11177087.0, + "step": 1547 + }, + { + "epoch": 1.5779816513761467, + "grad_norm": 5.6356964111328125, + "learning_rate": 1e-06, + "loss": 0.3589, + "mean_token_accuracy": 0.8578236103057861, + "num_tokens": 11185427.0, + "step": 1548 + }, + { + "epoch": 1.579001019367992, + "grad_norm": 5.534388542175293, + "learning_rate": 1e-06, + "loss": 0.2876, + "mean_token_accuracy": 0.8847275376319885, + "num_tokens": 11192326.0, + "step": 1549 + }, + { + "epoch": 1.5800203873598369, + "grad_norm": 4.5403523445129395, + "learning_rate": 1e-06, + "loss": 0.3109, + "mean_token_accuracy": 0.8790076375007629, + "num_tokens": 11201568.0, + "step": 1550 + }, + { + "epoch": 1.581039755351682, + "grad_norm": 5.2891387939453125, + "learning_rate": 1e-06, + "loss": 0.3133, + "mean_token_accuracy": 0.8733022212982178, + "num_tokens": 11209139.0, + "step": 1551 + }, + { + "epoch": 1.582059123343527, + "grad_norm": 5.2914719581604, + "learning_rate": 1e-06, + "loss": 0.298, + "mean_token_accuracy": 0.890758752822876, + "num_tokens": 11215712.0, + "step": 1552 + }, + { + "epoch": 1.583078491335372, + "grad_norm": 6.133773326873779, + "learning_rate": 1e-06, + "loss": 0.3265, + "mean_token_accuracy": 0.8724240064620972, + "num_tokens": 11221881.0, + "step": 1553 + }, + { + "epoch": 1.5840978593272173, + "grad_norm": 6.327688694000244, + "learning_rate": 1e-06, + "loss": 0.3461, + "mean_token_accuracy": 0.8673198223114014, + "num_tokens": 11228234.0, + "step": 1554 + }, + { + "epoch": 1.5851172273190621, + "grad_norm": 5.971930027008057, + "learning_rate": 1e-06, + "loss": 0.321, + "mean_token_accuracy": 0.8783782720565796, + "num_tokens": 11235894.0, + "step": 1555 + }, + { + "epoch": 1.5861365953109072, + "grad_norm": 5.297499179840088, + "learning_rate": 1e-06, + "loss": 0.3248, + "mean_token_accuracy": 0.8720386028289795, + "num_tokens": 11243499.0, + "step": 1556 + }, + { + "epoch": 1.5871559633027523, + "grad_norm": 5.5813703536987305, + "learning_rate": 1e-06, + "loss": 0.3439, + "mean_token_accuracy": 0.8652379512786865, + "num_tokens": 11249915.0, + "step": 1557 + }, + { + "epoch": 1.5881753312945972, + "grad_norm": 6.4816765785217285, + "learning_rate": 1e-06, + "loss": 0.3563, + "mean_token_accuracy": 0.861469566822052, + "num_tokens": 11256624.0, + "step": 1558 + }, + { + "epoch": 1.5891946992864425, + "grad_norm": 5.323218822479248, + "learning_rate": 1e-06, + "loss": 0.3078, + "mean_token_accuracy": 0.8812696933746338, + "num_tokens": 11264661.0, + "step": 1559 + }, + { + "epoch": 1.5902140672782874, + "grad_norm": 5.673209190368652, + "learning_rate": 1e-06, + "loss": 0.3585, + "mean_token_accuracy": 0.859625518321991, + "num_tokens": 11271550.0, + "step": 1560 + }, + { + "epoch": 1.5912334352701325, + "grad_norm": 4.9532551765441895, + "learning_rate": 1e-06, + "loss": 0.273, + "mean_token_accuracy": 0.8825621008872986, + "num_tokens": 11278880.0, + "step": 1561 + }, + { + "epoch": 1.5922528032619776, + "grad_norm": 6.639345645904541, + "learning_rate": 1e-06, + "loss": 0.3589, + "mean_token_accuracy": 0.8601152896881104, + "num_tokens": 11284524.0, + "step": 1562 + }, + { + "epoch": 1.5932721712538225, + "grad_norm": 5.818161964416504, + "learning_rate": 1e-06, + "loss": 0.3322, + "mean_token_accuracy": 0.8722974061965942, + "num_tokens": 11290666.0, + "step": 1563 + }, + { + "epoch": 1.5942915392456678, + "grad_norm": 5.578826904296875, + "learning_rate": 1e-06, + "loss": 0.3465, + "mean_token_accuracy": 0.861069917678833, + "num_tokens": 11297540.0, + "step": 1564 + }, + { + "epoch": 1.5953109072375127, + "grad_norm": 4.896342754364014, + "learning_rate": 1e-06, + "loss": 0.2833, + "mean_token_accuracy": 0.8894462585449219, + "num_tokens": 11305260.0, + "step": 1565 + }, + { + "epoch": 1.5963302752293578, + "grad_norm": 7.7351603507995605, + "learning_rate": 1e-06, + "loss": 0.3086, + "mean_token_accuracy": 0.884981632232666, + "num_tokens": 11313374.0, + "step": 1566 + }, + { + "epoch": 1.5973496432212029, + "grad_norm": 6.779548645019531, + "learning_rate": 1e-06, + "loss": 0.3734, + "mean_token_accuracy": 0.8534919023513794, + "num_tokens": 11318993.0, + "step": 1567 + }, + { + "epoch": 1.5983690112130478, + "grad_norm": 6.560418605804443, + "learning_rate": 1e-06, + "loss": 0.304, + "mean_token_accuracy": 0.8762904405593872, + "num_tokens": 11325070.0, + "step": 1568 + }, + { + "epoch": 1.599388379204893, + "grad_norm": 5.290040969848633, + "learning_rate": 1e-06, + "loss": 0.2996, + "mean_token_accuracy": 0.8799843788146973, + "num_tokens": 11333149.0, + "step": 1569 + }, + { + "epoch": 1.600407747196738, + "grad_norm": 5.963405132293701, + "learning_rate": 1e-06, + "loss": 0.3304, + "mean_token_accuracy": 0.8770018815994263, + "num_tokens": 11340321.0, + "step": 1570 + }, + { + "epoch": 1.601427115188583, + "grad_norm": 5.933952331542969, + "learning_rate": 1e-06, + "loss": 0.3287, + "mean_token_accuracy": 0.8742854595184326, + "num_tokens": 11347901.0, + "step": 1571 + }, + { + "epoch": 1.6024464831804281, + "grad_norm": 5.5948028564453125, + "learning_rate": 1e-06, + "loss": 0.3137, + "mean_token_accuracy": 0.8780736923217773, + "num_tokens": 11354741.0, + "step": 1572 + }, + { + "epoch": 1.603465851172273, + "grad_norm": 5.188351631164551, + "learning_rate": 1e-06, + "loss": 0.3064, + "mean_token_accuracy": 0.8819234371185303, + "num_tokens": 11361639.0, + "step": 1573 + }, + { + "epoch": 1.6044852191641183, + "grad_norm": 5.564644813537598, + "learning_rate": 1e-06, + "loss": 0.3154, + "mean_token_accuracy": 0.8725168704986572, + "num_tokens": 11369071.0, + "step": 1574 + }, + { + "epoch": 1.6055045871559632, + "grad_norm": 4.752492904663086, + "learning_rate": 1e-06, + "loss": 0.2927, + "mean_token_accuracy": 0.8847196698188782, + "num_tokens": 11377728.0, + "step": 1575 + }, + { + "epoch": 1.6065239551478083, + "grad_norm": 5.941027641296387, + "learning_rate": 1e-06, + "loss": 0.3242, + "mean_token_accuracy": 0.873573899269104, + "num_tokens": 11384104.0, + "step": 1576 + }, + { + "epoch": 1.6075433231396534, + "grad_norm": 6.418471813201904, + "learning_rate": 1e-06, + "loss": 0.3215, + "mean_token_accuracy": 0.8780486583709717, + "num_tokens": 11390279.0, + "step": 1577 + }, + { + "epoch": 1.6085626911314985, + "grad_norm": 7.942171096801758, + "learning_rate": 1e-06, + "loss": 0.3749, + "mean_token_accuracy": 0.8486348986625671, + "num_tokens": 11395549.0, + "step": 1578 + }, + { + "epoch": 1.6095820591233436, + "grad_norm": 5.159870624542236, + "learning_rate": 1e-06, + "loss": 0.3026, + "mean_token_accuracy": 0.8824747800827026, + "num_tokens": 11403261.0, + "step": 1579 + }, + { + "epoch": 1.6106014271151885, + "grad_norm": 5.520437717437744, + "learning_rate": 1e-06, + "loss": 0.2764, + "mean_token_accuracy": 0.8867914080619812, + "num_tokens": 11411322.0, + "step": 1580 + }, + { + "epoch": 1.6116207951070336, + "grad_norm": 5.113759517669678, + "learning_rate": 1e-06, + "loss": 0.2936, + "mean_token_accuracy": 0.8809515237808228, + "num_tokens": 11418772.0, + "step": 1581 + }, + { + "epoch": 1.6126401630988787, + "grad_norm": 4.417844295501709, + "learning_rate": 1e-06, + "loss": 0.2362, + "mean_token_accuracy": 0.9028599858283997, + "num_tokens": 11427652.0, + "step": 1582 + }, + { + "epoch": 1.6136595310907238, + "grad_norm": 5.355958461761475, + "learning_rate": 1e-06, + "loss": 0.3078, + "mean_token_accuracy": 0.8806302547454834, + "num_tokens": 11435519.0, + "step": 1583 + }, + { + "epoch": 1.614678899082569, + "grad_norm": 5.292672157287598, + "learning_rate": 1e-06, + "loss": 0.3047, + "mean_token_accuracy": 0.876502275466919, + "num_tokens": 11442100.0, + "step": 1584 + }, + { + "epoch": 1.6156982670744138, + "grad_norm": 5.721073627471924, + "learning_rate": 1e-06, + "loss": 0.3555, + "mean_token_accuracy": 0.8615880608558655, + "num_tokens": 11450032.0, + "step": 1585 + }, + { + "epoch": 1.6167176350662589, + "grad_norm": 5.0508131980896, + "learning_rate": 1e-06, + "loss": 0.286, + "mean_token_accuracy": 0.8816205859184265, + "num_tokens": 11457452.0, + "step": 1586 + }, + { + "epoch": 1.617737003058104, + "grad_norm": 5.297702789306641, + "learning_rate": 1e-06, + "loss": 0.3214, + "mean_token_accuracy": 0.8699486255645752, + "num_tokens": 11465283.0, + "step": 1587 + }, + { + "epoch": 1.618756371049949, + "grad_norm": 5.637813091278076, + "learning_rate": 1e-06, + "loss": 0.3342, + "mean_token_accuracy": 0.8720110654830933, + "num_tokens": 11473094.0, + "step": 1588 + }, + { + "epoch": 1.6197757390417942, + "grad_norm": 6.598003387451172, + "learning_rate": 1e-06, + "loss": 0.3324, + "mean_token_accuracy": 0.8656437993049622, + "num_tokens": 11479966.0, + "step": 1589 + }, + { + "epoch": 1.620795107033639, + "grad_norm": 4.922256946563721, + "learning_rate": 1e-06, + "loss": 0.2982, + "mean_token_accuracy": 0.8855921030044556, + "num_tokens": 11489534.0, + "step": 1590 + }, + { + "epoch": 1.6218144750254841, + "grad_norm": 4.9607110023498535, + "learning_rate": 1e-06, + "loss": 0.304, + "mean_token_accuracy": 0.8743104934692383, + "num_tokens": 11497434.0, + "step": 1591 + }, + { + "epoch": 1.6228338430173292, + "grad_norm": 5.3341755867004395, + "learning_rate": 1e-06, + "loss": 0.3084, + "mean_token_accuracy": 0.8739067316055298, + "num_tokens": 11504831.0, + "step": 1592 + }, + { + "epoch": 1.6238532110091743, + "grad_norm": 5.015634536743164, + "learning_rate": 1e-06, + "loss": 0.309, + "mean_token_accuracy": 0.8754328489303589, + "num_tokens": 11512895.0, + "step": 1593 + }, + { + "epoch": 1.6248725790010194, + "grad_norm": 6.310749530792236, + "learning_rate": 1e-06, + "loss": 0.3352, + "mean_token_accuracy": 0.8684455156326294, + "num_tokens": 11518500.0, + "step": 1594 + }, + { + "epoch": 1.6258919469928643, + "grad_norm": 5.620387554168701, + "learning_rate": 1e-06, + "loss": 0.3127, + "mean_token_accuracy": 0.8663701415061951, + "num_tokens": 11525042.0, + "step": 1595 + }, + { + "epoch": 1.6269113149847096, + "grad_norm": 5.597471714019775, + "learning_rate": 1e-06, + "loss": 0.3255, + "mean_token_accuracy": 0.8768125772476196, + "num_tokens": 11532364.0, + "step": 1596 + }, + { + "epoch": 1.6279306829765545, + "grad_norm": 4.665563583374023, + "learning_rate": 1e-06, + "loss": 0.2894, + "mean_token_accuracy": 0.8896266222000122, + "num_tokens": 11540279.0, + "step": 1597 + }, + { + "epoch": 1.6289500509683996, + "grad_norm": 4.606479644775391, + "learning_rate": 1e-06, + "loss": 0.2508, + "mean_token_accuracy": 0.9025169610977173, + "num_tokens": 11549940.0, + "step": 1598 + }, + { + "epoch": 1.6299694189602447, + "grad_norm": 5.330252647399902, + "learning_rate": 1e-06, + "loss": 0.3111, + "mean_token_accuracy": 0.873773455619812, + "num_tokens": 11556805.0, + "step": 1599 + }, + { + "epoch": 1.6309887869520896, + "grad_norm": 5.875285625457764, + "learning_rate": 1e-06, + "loss": 0.3435, + "mean_token_accuracy": 0.8708269000053406, + "num_tokens": 11563686.0, + "step": 1600 + }, + { + "epoch": 1.632008154943935, + "grad_norm": 5.806339263916016, + "learning_rate": 1e-06, + "loss": 0.3101, + "mean_token_accuracy": 0.8736018538475037, + "num_tokens": 11570575.0, + "step": 1601 + }, + { + "epoch": 1.6330275229357798, + "grad_norm": 6.431399345397949, + "learning_rate": 1e-06, + "loss": 0.3039, + "mean_token_accuracy": 0.8815670609474182, + "num_tokens": 11576910.0, + "step": 1602 + }, + { + "epoch": 1.634046890927625, + "grad_norm": 5.208342552185059, + "learning_rate": 1e-06, + "loss": 0.2952, + "mean_token_accuracy": 0.8878200054168701, + "num_tokens": 11583809.0, + "step": 1603 + }, + { + "epoch": 1.63506625891947, + "grad_norm": 5.560502529144287, + "learning_rate": 1e-06, + "loss": 0.3145, + "mean_token_accuracy": 0.8804234266281128, + "num_tokens": 11590249.0, + "step": 1604 + }, + { + "epoch": 1.6360856269113149, + "grad_norm": 6.00156307220459, + "learning_rate": 1e-06, + "loss": 0.3728, + "mean_token_accuracy": 0.8531547784805298, + "num_tokens": 11596324.0, + "step": 1605 + }, + { + "epoch": 1.6371049949031602, + "grad_norm": 5.364150524139404, + "learning_rate": 1e-06, + "loss": 0.3118, + "mean_token_accuracy": 0.8785151839256287, + "num_tokens": 11603430.0, + "step": 1606 + }, + { + "epoch": 1.638124362895005, + "grad_norm": 5.292736053466797, + "learning_rate": 1e-06, + "loss": 0.3106, + "mean_token_accuracy": 0.870209276676178, + "num_tokens": 11609844.0, + "step": 1607 + }, + { + "epoch": 1.6391437308868502, + "grad_norm": 5.097550392150879, + "learning_rate": 1e-06, + "loss": 0.334, + "mean_token_accuracy": 0.8740018606185913, + "num_tokens": 11618296.0, + "step": 1608 + }, + { + "epoch": 1.6401630988786953, + "grad_norm": 5.326269626617432, + "learning_rate": 1e-06, + "loss": 0.3491, + "mean_token_accuracy": 0.8672267198562622, + "num_tokens": 11625196.0, + "step": 1609 + }, + { + "epoch": 1.6411824668705401, + "grad_norm": 6.518932342529297, + "learning_rate": 1e-06, + "loss": 0.3136, + "mean_token_accuracy": 0.8796922564506531, + "num_tokens": 11631135.0, + "step": 1610 + }, + { + "epoch": 1.6422018348623855, + "grad_norm": 6.021076679229736, + "learning_rate": 1e-06, + "loss": 0.336, + "mean_token_accuracy": 0.8659901022911072, + "num_tokens": 11637451.0, + "step": 1611 + }, + { + "epoch": 1.6432212028542303, + "grad_norm": 4.879341125488281, + "learning_rate": 1e-06, + "loss": 0.2929, + "mean_token_accuracy": 0.8865424394607544, + "num_tokens": 11645219.0, + "step": 1612 + }, + { + "epoch": 1.6442405708460754, + "grad_norm": 4.776038646697998, + "learning_rate": 1e-06, + "loss": 0.3024, + "mean_token_accuracy": 0.8816410303115845, + "num_tokens": 11653593.0, + "step": 1613 + }, + { + "epoch": 1.6452599388379205, + "grad_norm": 5.898937225341797, + "learning_rate": 1e-06, + "loss": 0.3136, + "mean_token_accuracy": 0.8780398368835449, + "num_tokens": 11660920.0, + "step": 1614 + }, + { + "epoch": 1.6462793068297654, + "grad_norm": 5.685494899749756, + "learning_rate": 1e-06, + "loss": 0.3279, + "mean_token_accuracy": 0.8664219379425049, + "num_tokens": 11668222.0, + "step": 1615 + }, + { + "epoch": 1.6472986748216107, + "grad_norm": 6.448815822601318, + "learning_rate": 1e-06, + "loss": 0.339, + "mean_token_accuracy": 0.8662570714950562, + "num_tokens": 11674027.0, + "step": 1616 + }, + { + "epoch": 1.6483180428134556, + "grad_norm": 5.438894271850586, + "learning_rate": 1e-06, + "loss": 0.3184, + "mean_token_accuracy": 0.8706713914871216, + "num_tokens": 11681286.0, + "step": 1617 + }, + { + "epoch": 1.6493374108053007, + "grad_norm": 5.14253568649292, + "learning_rate": 1e-06, + "loss": 0.2834, + "mean_token_accuracy": 0.8828875422477722, + "num_tokens": 11688494.0, + "step": 1618 + }, + { + "epoch": 1.6503567787971458, + "grad_norm": 4.842401027679443, + "learning_rate": 1e-06, + "loss": 0.2865, + "mean_token_accuracy": 0.8894822001457214, + "num_tokens": 11696590.0, + "step": 1619 + }, + { + "epoch": 1.6513761467889907, + "grad_norm": 5.692100524902344, + "learning_rate": 1e-06, + "loss": 0.3283, + "mean_token_accuracy": 0.8733252286911011, + "num_tokens": 11703901.0, + "step": 1620 + }, + { + "epoch": 1.652395514780836, + "grad_norm": 4.9341044425964355, + "learning_rate": 1e-06, + "loss": 0.2679, + "mean_token_accuracy": 0.8971724510192871, + "num_tokens": 11712278.0, + "step": 1621 + }, + { + "epoch": 1.653414882772681, + "grad_norm": 4.935537815093994, + "learning_rate": 1e-06, + "loss": 0.3143, + "mean_token_accuracy": 0.8669471144676208, + "num_tokens": 11720801.0, + "step": 1622 + }, + { + "epoch": 1.654434250764526, + "grad_norm": 6.319352626800537, + "learning_rate": 1e-06, + "loss": 0.3558, + "mean_token_accuracy": 0.8603435754776001, + "num_tokens": 11727647.0, + "step": 1623 + }, + { + "epoch": 1.655453618756371, + "grad_norm": 5.200064182281494, + "learning_rate": 1e-06, + "loss": 0.274, + "mean_token_accuracy": 0.8707215785980225, + "num_tokens": 11735010.0, + "step": 1624 + }, + { + "epoch": 1.656472986748216, + "grad_norm": 5.412961483001709, + "learning_rate": 1e-06, + "loss": 0.3279, + "mean_token_accuracy": 0.8699284791946411, + "num_tokens": 11741894.0, + "step": 1625 + }, + { + "epoch": 1.6574923547400613, + "grad_norm": 5.503965377807617, + "learning_rate": 1e-06, + "loss": 0.347, + "mean_token_accuracy": 0.8724486827850342, + "num_tokens": 11750490.0, + "step": 1626 + }, + { + "epoch": 1.6585117227319062, + "grad_norm": 4.767349720001221, + "learning_rate": 1e-06, + "loss": 0.3042, + "mean_token_accuracy": 0.8786114454269409, + "num_tokens": 11759178.0, + "step": 1627 + }, + { + "epoch": 1.6595310907237513, + "grad_norm": 5.401465892791748, + "learning_rate": 1e-06, + "loss": 0.3168, + "mean_token_accuracy": 0.8791537880897522, + "num_tokens": 11766506.0, + "step": 1628 + }, + { + "epoch": 1.6605504587155964, + "grad_norm": 3.958631992340088, + "learning_rate": 1e-06, + "loss": 0.2096, + "mean_token_accuracy": 0.9076091647148132, + "num_tokens": 11775517.0, + "step": 1629 + }, + { + "epoch": 1.6615698267074412, + "grad_norm": 4.496771812438965, + "learning_rate": 1e-06, + "loss": 0.3046, + "mean_token_accuracy": 0.8852639198303223, + "num_tokens": 11784737.0, + "step": 1630 + }, + { + "epoch": 1.6625891946992866, + "grad_norm": 4.495499134063721, + "learning_rate": 1e-06, + "loss": 0.2605, + "mean_token_accuracy": 0.8987916707992554, + "num_tokens": 11794071.0, + "step": 1631 + }, + { + "epoch": 1.6636085626911314, + "grad_norm": 5.316571235656738, + "learning_rate": 1e-06, + "loss": 0.3237, + "mean_token_accuracy": 0.863568902015686, + "num_tokens": 11801452.0, + "step": 1632 + }, + { + "epoch": 1.6646279306829765, + "grad_norm": 4.651771545410156, + "learning_rate": 1e-06, + "loss": 0.2842, + "mean_token_accuracy": 0.8923529982566833, + "num_tokens": 11809252.0, + "step": 1633 + }, + { + "epoch": 1.6656472986748216, + "grad_norm": 6.067620754241943, + "learning_rate": 1e-06, + "loss": 0.3228, + "mean_token_accuracy": 0.8728115558624268, + "num_tokens": 11816541.0, + "step": 1634 + }, + { + "epoch": 1.6666666666666665, + "grad_norm": 5.039408206939697, + "learning_rate": 1e-06, + "loss": 0.2969, + "mean_token_accuracy": 0.8826560974121094, + "num_tokens": 11824018.0, + "step": 1635 + }, + { + "epoch": 1.6676860346585118, + "grad_norm": 5.123334884643555, + "learning_rate": 1e-06, + "loss": 0.305, + "mean_token_accuracy": 0.8815345764160156, + "num_tokens": 11832285.0, + "step": 1636 + }, + { + "epoch": 1.6687054026503567, + "grad_norm": 4.998981952667236, + "learning_rate": 1e-06, + "loss": 0.3053, + "mean_token_accuracy": 0.8830667734146118, + "num_tokens": 11840153.0, + "step": 1637 + }, + { + "epoch": 1.6697247706422018, + "grad_norm": 5.536434173583984, + "learning_rate": 1e-06, + "loss": 0.3027, + "mean_token_accuracy": 0.8797615766525269, + "num_tokens": 11847438.0, + "step": 1638 + }, + { + "epoch": 1.670744138634047, + "grad_norm": 6.266422748565674, + "learning_rate": 1e-06, + "loss": 0.3176, + "mean_token_accuracy": 0.8764654994010925, + "num_tokens": 11853653.0, + "step": 1639 + }, + { + "epoch": 1.6717635066258918, + "grad_norm": 5.965389251708984, + "learning_rate": 1e-06, + "loss": 0.3195, + "mean_token_accuracy": 0.8778331279754639, + "num_tokens": 11859732.0, + "step": 1640 + }, + { + "epoch": 1.6727828746177371, + "grad_norm": 5.220846652984619, + "learning_rate": 1e-06, + "loss": 0.3104, + "mean_token_accuracy": 0.883171796798706, + "num_tokens": 11867356.0, + "step": 1641 + }, + { + "epoch": 1.673802242609582, + "grad_norm": 5.5492844581604, + "learning_rate": 1e-06, + "loss": 0.3132, + "mean_token_accuracy": 0.8673964738845825, + "num_tokens": 11874091.0, + "step": 1642 + }, + { + "epoch": 1.674821610601427, + "grad_norm": 5.390706539154053, + "learning_rate": 1e-06, + "loss": 0.3072, + "mean_token_accuracy": 0.8811319470405579, + "num_tokens": 11881680.0, + "step": 1643 + }, + { + "epoch": 1.6758409785932722, + "grad_norm": 5.786391258239746, + "learning_rate": 1e-06, + "loss": 0.3464, + "mean_token_accuracy": 0.8591821193695068, + "num_tokens": 11887840.0, + "step": 1644 + }, + { + "epoch": 1.676860346585117, + "grad_norm": 4.7877583503723145, + "learning_rate": 1e-06, + "loss": 0.2959, + "mean_token_accuracy": 0.8808383941650391, + "num_tokens": 11896367.0, + "step": 1645 + }, + { + "epoch": 1.6778797145769624, + "grad_norm": 4.337416648864746, + "learning_rate": 1e-06, + "loss": 0.2749, + "mean_token_accuracy": 0.8849438428878784, + "num_tokens": 11904977.0, + "step": 1646 + }, + { + "epoch": 1.6788990825688073, + "grad_norm": 5.291985511779785, + "learning_rate": 1e-06, + "loss": 0.2954, + "mean_token_accuracy": 0.8745111227035522, + "num_tokens": 11912153.0, + "step": 1647 + }, + { + "epoch": 1.6799184505606524, + "grad_norm": 5.766292095184326, + "learning_rate": 1e-06, + "loss": 0.3181, + "mean_token_accuracy": 0.8702427744865417, + "num_tokens": 11919260.0, + "step": 1648 + }, + { + "epoch": 1.6809378185524975, + "grad_norm": 4.6046528816223145, + "learning_rate": 1e-06, + "loss": 0.3187, + "mean_token_accuracy": 0.878964900970459, + "num_tokens": 11929664.0, + "step": 1649 + }, + { + "epoch": 1.6819571865443423, + "grad_norm": 5.51922607421875, + "learning_rate": 1e-06, + "loss": 0.3196, + "mean_token_accuracy": 0.8745816946029663, + "num_tokens": 11936676.0, + "step": 1650 + }, + { + "epoch": 1.6829765545361877, + "grad_norm": 5.737668991088867, + "learning_rate": 1e-06, + "loss": 0.3312, + "mean_token_accuracy": 0.8690133094787598, + "num_tokens": 11942343.0, + "step": 1651 + }, + { + "epoch": 1.6839959225280325, + "grad_norm": 6.483734130859375, + "learning_rate": 1e-06, + "loss": 0.3398, + "mean_token_accuracy": 0.8695509433746338, + "num_tokens": 11947678.0, + "step": 1652 + }, + { + "epoch": 1.6850152905198776, + "grad_norm": 4.819568157196045, + "learning_rate": 1e-06, + "loss": 0.2777, + "mean_token_accuracy": 0.8906035423278809, + "num_tokens": 11955257.0, + "step": 1653 + }, + { + "epoch": 1.6860346585117227, + "grad_norm": 5.490198612213135, + "learning_rate": 1e-06, + "loss": 0.3477, + "mean_token_accuracy": 0.862739086151123, + "num_tokens": 11962388.0, + "step": 1654 + }, + { + "epoch": 1.6870540265035678, + "grad_norm": 5.141986846923828, + "learning_rate": 1e-06, + "loss": 0.2769, + "mean_token_accuracy": 0.8893186450004578, + "num_tokens": 11969129.0, + "step": 1655 + }, + { + "epoch": 1.688073394495413, + "grad_norm": 5.34635591506958, + "learning_rate": 1e-06, + "loss": 0.3121, + "mean_token_accuracy": 0.8792014122009277, + "num_tokens": 11976163.0, + "step": 1656 + }, + { + "epoch": 1.6890927624872578, + "grad_norm": 5.152719974517822, + "learning_rate": 1e-06, + "loss": 0.3013, + "mean_token_accuracy": 0.8851742744445801, + "num_tokens": 11983937.0, + "step": 1657 + }, + { + "epoch": 1.690112130479103, + "grad_norm": 5.7441205978393555, + "learning_rate": 1e-06, + "loss": 0.3212, + "mean_token_accuracy": 0.8783098459243774, + "num_tokens": 11991164.0, + "step": 1658 + }, + { + "epoch": 1.691131498470948, + "grad_norm": 5.925512790679932, + "learning_rate": 1e-06, + "loss": 0.3142, + "mean_token_accuracy": 0.8740756511688232, + "num_tokens": 11997042.0, + "step": 1659 + }, + { + "epoch": 1.6921508664627931, + "grad_norm": 5.281119346618652, + "learning_rate": 1e-06, + "loss": 0.299, + "mean_token_accuracy": 0.872144341468811, + "num_tokens": 12005211.0, + "step": 1660 + }, + { + "epoch": 1.6931702344546382, + "grad_norm": 5.953085422515869, + "learning_rate": 1e-06, + "loss": 0.3147, + "mean_token_accuracy": 0.8726950883865356, + "num_tokens": 12011505.0, + "step": 1661 + }, + { + "epoch": 1.694189602446483, + "grad_norm": 4.726206302642822, + "learning_rate": 1e-06, + "loss": 0.2928, + "mean_token_accuracy": 0.8820914626121521, + "num_tokens": 12019380.0, + "step": 1662 + }, + { + "epoch": 1.6952089704383282, + "grad_norm": 5.803210258483887, + "learning_rate": 1e-06, + "loss": 0.3218, + "mean_token_accuracy": 0.8744301795959473, + "num_tokens": 12025268.0, + "step": 1663 + }, + { + "epoch": 1.6962283384301733, + "grad_norm": 6.884122848510742, + "learning_rate": 1e-06, + "loss": 0.3156, + "mean_token_accuracy": 0.8833150863647461, + "num_tokens": 12030829.0, + "step": 1664 + }, + { + "epoch": 1.6972477064220184, + "grad_norm": 5.3438544273376465, + "learning_rate": 1e-06, + "loss": 0.324, + "mean_token_accuracy": 0.8685641884803772, + "num_tokens": 12038275.0, + "step": 1665 + }, + { + "epoch": 1.6982670744138635, + "grad_norm": 6.784457683563232, + "learning_rate": 1e-06, + "loss": 0.3392, + "mean_token_accuracy": 0.865423321723938, + "num_tokens": 12044810.0, + "step": 1666 + }, + { + "epoch": 1.6992864424057084, + "grad_norm": 4.512589931488037, + "learning_rate": 1e-06, + "loss": 0.2593, + "mean_token_accuracy": 0.8893846869468689, + "num_tokens": 12053413.0, + "step": 1667 + }, + { + "epoch": 1.7003058103975535, + "grad_norm": 5.894179821014404, + "learning_rate": 1e-06, + "loss": 0.3094, + "mean_token_accuracy": 0.8801732063293457, + "num_tokens": 12060292.0, + "step": 1668 + }, + { + "epoch": 1.7013251783893986, + "grad_norm": 5.174228668212891, + "learning_rate": 1e-06, + "loss": 0.3182, + "mean_token_accuracy": 0.8814761638641357, + "num_tokens": 12069305.0, + "step": 1669 + }, + { + "epoch": 1.7023445463812437, + "grad_norm": 7.11729621887207, + "learning_rate": 1e-06, + "loss": 0.3781, + "mean_token_accuracy": 0.846440851688385, + "num_tokens": 12074652.0, + "step": 1670 + }, + { + "epoch": 1.7033639143730888, + "grad_norm": 5.8631591796875, + "learning_rate": 1e-06, + "loss": 0.3239, + "mean_token_accuracy": 0.8713518381118774, + "num_tokens": 12081565.0, + "step": 1671 + }, + { + "epoch": 1.7043832823649336, + "grad_norm": 5.404301643371582, + "learning_rate": 1e-06, + "loss": 0.3199, + "mean_token_accuracy": 0.8763270378112793, + "num_tokens": 12088386.0, + "step": 1672 + }, + { + "epoch": 1.705402650356779, + "grad_norm": 5.578619956970215, + "learning_rate": 1e-06, + "loss": 0.337, + "mean_token_accuracy": 0.8698873519897461, + "num_tokens": 12095018.0, + "step": 1673 + }, + { + "epoch": 1.7064220183486238, + "grad_norm": 4.805509567260742, + "learning_rate": 1e-06, + "loss": 0.3115, + "mean_token_accuracy": 0.8800532817840576, + "num_tokens": 12103483.0, + "step": 1674 + }, + { + "epoch": 1.707441386340469, + "grad_norm": 5.14913272857666, + "learning_rate": 1e-06, + "loss": 0.311, + "mean_token_accuracy": 0.8788926601409912, + "num_tokens": 12111183.0, + "step": 1675 + }, + { + "epoch": 1.708460754332314, + "grad_norm": 5.590290546417236, + "learning_rate": 1e-06, + "loss": 0.2893, + "mean_token_accuracy": 0.8849366307258606, + "num_tokens": 12117372.0, + "step": 1676 + }, + { + "epoch": 1.709480122324159, + "grad_norm": 5.919790267944336, + "learning_rate": 1e-06, + "loss": 0.3428, + "mean_token_accuracy": 0.878820538520813, + "num_tokens": 12124780.0, + "step": 1677 + }, + { + "epoch": 1.7104994903160042, + "grad_norm": 5.158208847045898, + "learning_rate": 1e-06, + "loss": 0.3051, + "mean_token_accuracy": 0.8815770745277405, + "num_tokens": 12132816.0, + "step": 1678 + }, + { + "epoch": 1.7115188583078491, + "grad_norm": 5.4693379402160645, + "learning_rate": 1e-06, + "loss": 0.3347, + "mean_token_accuracy": 0.8664238452911377, + "num_tokens": 12139527.0, + "step": 1679 + }, + { + "epoch": 1.7125382262996942, + "grad_norm": 6.021618843078613, + "learning_rate": 1e-06, + "loss": 0.3218, + "mean_token_accuracy": 0.875586748123169, + "num_tokens": 12147019.0, + "step": 1680 + }, + { + "epoch": 1.7135575942915393, + "grad_norm": 5.041711807250977, + "learning_rate": 1e-06, + "loss": 0.2867, + "mean_token_accuracy": 0.888920247554779, + "num_tokens": 12154507.0, + "step": 1681 + }, + { + "epoch": 1.7145769622833842, + "grad_norm": 5.82397985458374, + "learning_rate": 1e-06, + "loss": 0.3462, + "mean_token_accuracy": 0.8712193369865417, + "num_tokens": 12161462.0, + "step": 1682 + }, + { + "epoch": 1.7155963302752295, + "grad_norm": 6.000777721405029, + "learning_rate": 1e-06, + "loss": 0.3436, + "mean_token_accuracy": 0.8590835332870483, + "num_tokens": 12169058.0, + "step": 1683 + }, + { + "epoch": 1.7166156982670744, + "grad_norm": 5.496016502380371, + "learning_rate": 1e-06, + "loss": 0.2965, + "mean_token_accuracy": 0.8800785541534424, + "num_tokens": 12176042.0, + "step": 1684 + }, + { + "epoch": 1.7176350662589195, + "grad_norm": 5.615152359008789, + "learning_rate": 1e-06, + "loss": 0.3084, + "mean_token_accuracy": 0.8810091614723206, + "num_tokens": 12182120.0, + "step": 1685 + }, + { + "epoch": 1.7186544342507646, + "grad_norm": 5.794187068939209, + "learning_rate": 1e-06, + "loss": 0.3304, + "mean_token_accuracy": 0.8755162358283997, + "num_tokens": 12189411.0, + "step": 1686 + }, + { + "epoch": 1.7196738022426095, + "grad_norm": 4.629827976226807, + "learning_rate": 1e-06, + "loss": 0.2727, + "mean_token_accuracy": 0.8969591856002808, + "num_tokens": 12197255.0, + "step": 1687 + }, + { + "epoch": 1.7206931702344548, + "grad_norm": 5.786254405975342, + "learning_rate": 1e-06, + "loss": 0.3121, + "mean_token_accuracy": 0.87364262342453, + "num_tokens": 12203761.0, + "step": 1688 + }, + { + "epoch": 1.7217125382262997, + "grad_norm": 5.77506685256958, + "learning_rate": 1e-06, + "loss": 0.3423, + "mean_token_accuracy": 0.8697848320007324, + "num_tokens": 12210503.0, + "step": 1689 + }, + { + "epoch": 1.7227319062181448, + "grad_norm": 6.100305080413818, + "learning_rate": 1e-06, + "loss": 0.303, + "mean_token_accuracy": 0.881568968296051, + "num_tokens": 12216991.0, + "step": 1690 + }, + { + "epoch": 1.7237512742099899, + "grad_norm": 5.611420154571533, + "learning_rate": 1e-06, + "loss": 0.3088, + "mean_token_accuracy": 0.8704565763473511, + "num_tokens": 12223495.0, + "step": 1691 + }, + { + "epoch": 1.7247706422018347, + "grad_norm": 5.940615177154541, + "learning_rate": 1e-06, + "loss": 0.315, + "mean_token_accuracy": 0.8820200562477112, + "num_tokens": 12230350.0, + "step": 1692 + }, + { + "epoch": 1.72579001019368, + "grad_norm": 5.3119215965271, + "learning_rate": 1e-06, + "loss": 0.3109, + "mean_token_accuracy": 0.8807504177093506, + "num_tokens": 12237373.0, + "step": 1693 + }, + { + "epoch": 1.726809378185525, + "grad_norm": 4.808427333831787, + "learning_rate": 1e-06, + "loss": 0.3076, + "mean_token_accuracy": 0.8801932334899902, + "num_tokens": 12246446.0, + "step": 1694 + }, + { + "epoch": 1.72782874617737, + "grad_norm": 5.921219348907471, + "learning_rate": 1e-06, + "loss": 0.3686, + "mean_token_accuracy": 0.8563852310180664, + "num_tokens": 12253196.0, + "step": 1695 + }, + { + "epoch": 1.7288481141692151, + "grad_norm": 5.021627902984619, + "learning_rate": 1e-06, + "loss": 0.2834, + "mean_token_accuracy": 0.8860157132148743, + "num_tokens": 12261064.0, + "step": 1696 + }, + { + "epoch": 1.72986748216106, + "grad_norm": 5.75499153137207, + "learning_rate": 1e-06, + "loss": 0.3266, + "mean_token_accuracy": 0.8769583702087402, + "num_tokens": 12267897.0, + "step": 1697 + }, + { + "epoch": 1.7308868501529053, + "grad_norm": 5.392025947570801, + "learning_rate": 1e-06, + "loss": 0.3143, + "mean_token_accuracy": 0.8750849962234497, + "num_tokens": 12274870.0, + "step": 1698 + }, + { + "epoch": 1.7319062181447502, + "grad_norm": 5.809897422790527, + "learning_rate": 1e-06, + "loss": 0.344, + "mean_token_accuracy": 0.8675345182418823, + "num_tokens": 12281923.0, + "step": 1699 + }, + { + "epoch": 1.7329255861365953, + "grad_norm": 5.1436543464660645, + "learning_rate": 1e-06, + "loss": 0.3054, + "mean_token_accuracy": 0.8826211094856262, + "num_tokens": 12289235.0, + "step": 1700 + }, + { + "epoch": 1.7339449541284404, + "grad_norm": 4.726084232330322, + "learning_rate": 1e-06, + "loss": 0.2886, + "mean_token_accuracy": 0.8874056339263916, + "num_tokens": 12297440.0, + "step": 1701 + }, + { + "epoch": 1.7349643221202853, + "grad_norm": 5.76243782043457, + "learning_rate": 1e-06, + "loss": 0.3462, + "mean_token_accuracy": 0.8606798648834229, + "num_tokens": 12303839.0, + "step": 1702 + }, + { + "epoch": 1.7359836901121306, + "grad_norm": 5.315376281738281, + "learning_rate": 1e-06, + "loss": 0.312, + "mean_token_accuracy": 0.8835283517837524, + "num_tokens": 12311066.0, + "step": 1703 + }, + { + "epoch": 1.7370030581039755, + "grad_norm": 5.485423564910889, + "learning_rate": 1e-06, + "loss": 0.3204, + "mean_token_accuracy": 0.8796330690383911, + "num_tokens": 12317998.0, + "step": 1704 + }, + { + "epoch": 1.7380224260958206, + "grad_norm": 4.977173328399658, + "learning_rate": 1e-06, + "loss": 0.2882, + "mean_token_accuracy": 0.8902087211608887, + "num_tokens": 12326451.0, + "step": 1705 + }, + { + "epoch": 1.7390417940876657, + "grad_norm": 5.5054931640625, + "learning_rate": 1e-06, + "loss": 0.3176, + "mean_token_accuracy": 0.8798881769180298, + "num_tokens": 12333888.0, + "step": 1706 + }, + { + "epoch": 1.7400611620795106, + "grad_norm": 5.660472393035889, + "learning_rate": 1e-06, + "loss": 0.3238, + "mean_token_accuracy": 0.8711731433868408, + "num_tokens": 12340179.0, + "step": 1707 + }, + { + "epoch": 1.7410805300713559, + "grad_norm": 5.831614017486572, + "learning_rate": 1e-06, + "loss": 0.3441, + "mean_token_accuracy": 0.8578923344612122, + "num_tokens": 12346261.0, + "step": 1708 + }, + { + "epoch": 1.7420998980632008, + "grad_norm": 6.065323829650879, + "learning_rate": 1e-06, + "loss": 0.3304, + "mean_token_accuracy": 0.8704382181167603, + "num_tokens": 12352702.0, + "step": 1709 + }, + { + "epoch": 1.7431192660550459, + "grad_norm": 4.811304569244385, + "learning_rate": 1e-06, + "loss": 0.2999, + "mean_token_accuracy": 0.8790035247802734, + "num_tokens": 12361357.0, + "step": 1710 + }, + { + "epoch": 1.744138634046891, + "grad_norm": 4.929521560668945, + "learning_rate": 1e-06, + "loss": 0.3098, + "mean_token_accuracy": 0.8717650175094604, + "num_tokens": 12369911.0, + "step": 1711 + }, + { + "epoch": 1.7451580020387358, + "grad_norm": 5.818469047546387, + "learning_rate": 1e-06, + "loss": 0.3309, + "mean_token_accuracy": 0.8669000864028931, + "num_tokens": 12376257.0, + "step": 1712 + }, + { + "epoch": 1.7461773700305812, + "grad_norm": 5.630898952484131, + "learning_rate": 1e-06, + "loss": 0.282, + "mean_token_accuracy": 0.8897846341133118, + "num_tokens": 12383479.0, + "step": 1713 + }, + { + "epoch": 1.747196738022426, + "grad_norm": 5.771079063415527, + "learning_rate": 1e-06, + "loss": 0.3179, + "mean_token_accuracy": 0.8674855828285217, + "num_tokens": 12390499.0, + "step": 1714 + }, + { + "epoch": 1.7482161060142711, + "grad_norm": 4.800970077514648, + "learning_rate": 1e-06, + "loss": 0.2769, + "mean_token_accuracy": 0.8910151720046997, + "num_tokens": 12399347.0, + "step": 1715 + }, + { + "epoch": 1.7492354740061162, + "grad_norm": 4.944757461547852, + "learning_rate": 1e-06, + "loss": 0.2802, + "mean_token_accuracy": 0.885566234588623, + "num_tokens": 12407565.0, + "step": 1716 + }, + { + "epoch": 1.750254841997961, + "grad_norm": 5.448134422302246, + "learning_rate": 1e-06, + "loss": 0.2997, + "mean_token_accuracy": 0.8812716603279114, + "num_tokens": 12414738.0, + "step": 1717 + }, + { + "epoch": 1.7512742099898064, + "grad_norm": 6.717865467071533, + "learning_rate": 1e-06, + "loss": 0.3302, + "mean_token_accuracy": 0.869944155216217, + "num_tokens": 12420337.0, + "step": 1718 + }, + { + "epoch": 1.7522935779816513, + "grad_norm": 5.426316738128662, + "learning_rate": 1e-06, + "loss": 0.2783, + "mean_token_accuracy": 0.8944811224937439, + "num_tokens": 12427441.0, + "step": 1719 + }, + { + "epoch": 1.7533129459734964, + "grad_norm": 4.6988396644592285, + "learning_rate": 1e-06, + "loss": 0.3046, + "mean_token_accuracy": 0.8767707347869873, + "num_tokens": 12435879.0, + "step": 1720 + }, + { + "epoch": 1.7543323139653415, + "grad_norm": 5.819405555725098, + "learning_rate": 1e-06, + "loss": 0.3443, + "mean_token_accuracy": 0.8680524826049805, + "num_tokens": 12442849.0, + "step": 1721 + }, + { + "epoch": 1.7553516819571864, + "grad_norm": 5.613722324371338, + "learning_rate": 1e-06, + "loss": 0.3059, + "mean_token_accuracy": 0.883230447769165, + "num_tokens": 12449095.0, + "step": 1722 + }, + { + "epoch": 1.7563710499490317, + "grad_norm": 4.90739631652832, + "learning_rate": 1e-06, + "loss": 0.3158, + "mean_token_accuracy": 0.8783197402954102, + "num_tokens": 12458491.0, + "step": 1723 + }, + { + "epoch": 1.7573904179408766, + "grad_norm": 5.8975300788879395, + "learning_rate": 1e-06, + "loss": 0.3421, + "mean_token_accuracy": 0.8580281138420105, + "num_tokens": 12465045.0, + "step": 1724 + }, + { + "epoch": 1.7584097859327217, + "grad_norm": 5.203415393829346, + "learning_rate": 1e-06, + "loss": 0.3373, + "mean_token_accuracy": 0.8613122701644897, + "num_tokens": 12472959.0, + "step": 1725 + }, + { + "epoch": 1.7594291539245668, + "grad_norm": 5.569844722747803, + "learning_rate": 1e-06, + "loss": 0.3211, + "mean_token_accuracy": 0.8703032732009888, + "num_tokens": 12480375.0, + "step": 1726 + }, + { + "epoch": 1.7604485219164119, + "grad_norm": 5.1015214920043945, + "learning_rate": 1e-06, + "loss": 0.3055, + "mean_token_accuracy": 0.877860426902771, + "num_tokens": 12487946.0, + "step": 1727 + }, + { + "epoch": 1.761467889908257, + "grad_norm": 5.578083038330078, + "learning_rate": 1e-06, + "loss": 0.3333, + "mean_token_accuracy": 0.8701976537704468, + "num_tokens": 12494683.0, + "step": 1728 + }, + { + "epoch": 1.7624872579001019, + "grad_norm": 4.5243821144104, + "learning_rate": 1e-06, + "loss": 0.2783, + "mean_token_accuracy": 0.8915731310844421, + "num_tokens": 12503488.0, + "step": 1729 + }, + { + "epoch": 1.763506625891947, + "grad_norm": 4.510097026824951, + "learning_rate": 1e-06, + "loss": 0.295, + "mean_token_accuracy": 0.8801817893981934, + "num_tokens": 12511319.0, + "step": 1730 + }, + { + "epoch": 1.764525993883792, + "grad_norm": 5.364851951599121, + "learning_rate": 1e-06, + "loss": 0.3405, + "mean_token_accuracy": 0.8681402206420898, + "num_tokens": 12518759.0, + "step": 1731 + }, + { + "epoch": 1.7655453618756372, + "grad_norm": 4.757434368133545, + "learning_rate": 1e-06, + "loss": 0.2964, + "mean_token_accuracy": 0.8802605867385864, + "num_tokens": 12526949.0, + "step": 1732 + }, + { + "epoch": 1.7665647298674823, + "grad_norm": 4.642997741699219, + "learning_rate": 1e-06, + "loss": 0.28, + "mean_token_accuracy": 0.8919385075569153, + "num_tokens": 12534978.0, + "step": 1733 + }, + { + "epoch": 1.7675840978593271, + "grad_norm": 4.425345420837402, + "learning_rate": 1e-06, + "loss": 0.2682, + "mean_token_accuracy": 0.8950656652450562, + "num_tokens": 12543419.0, + "step": 1734 + }, + { + "epoch": 1.7686034658511722, + "grad_norm": 5.287744045257568, + "learning_rate": 1e-06, + "loss": 0.3029, + "mean_token_accuracy": 0.8830204010009766, + "num_tokens": 12550240.0, + "step": 1735 + }, + { + "epoch": 1.7696228338430173, + "grad_norm": 4.818442344665527, + "learning_rate": 1e-06, + "loss": 0.3124, + "mean_token_accuracy": 0.8773698210716248, + "num_tokens": 12559040.0, + "step": 1736 + }, + { + "epoch": 1.7706422018348624, + "grad_norm": 5.384751796722412, + "learning_rate": 1e-06, + "loss": 0.3086, + "mean_token_accuracy": 0.8837053775787354, + "num_tokens": 12565514.0, + "step": 1737 + }, + { + "epoch": 1.7716615698267075, + "grad_norm": 5.971268653869629, + "learning_rate": 1e-06, + "loss": 0.3274, + "mean_token_accuracy": 0.8717244863510132, + "num_tokens": 12571917.0, + "step": 1738 + }, + { + "epoch": 1.7726809378185524, + "grad_norm": 5.59006404876709, + "learning_rate": 1e-06, + "loss": 0.3052, + "mean_token_accuracy": 0.8759239912033081, + "num_tokens": 12579429.0, + "step": 1739 + }, + { + "epoch": 1.7737003058103975, + "grad_norm": 4.8286824226379395, + "learning_rate": 1e-06, + "loss": 0.2817, + "mean_token_accuracy": 0.8884164094924927, + "num_tokens": 12587593.0, + "step": 1740 + }, + { + "epoch": 1.7747196738022426, + "grad_norm": 6.494954586029053, + "learning_rate": 1e-06, + "loss": 0.336, + "mean_token_accuracy": 0.86891770362854, + "num_tokens": 12594548.0, + "step": 1741 + }, + { + "epoch": 1.7757390417940877, + "grad_norm": 5.509841442108154, + "learning_rate": 1e-06, + "loss": 0.3327, + "mean_token_accuracy": 0.8681806325912476, + "num_tokens": 12600841.0, + "step": 1742 + }, + { + "epoch": 1.7767584097859328, + "grad_norm": 6.137204647064209, + "learning_rate": 1e-06, + "loss": 0.2869, + "mean_token_accuracy": 0.8852636814117432, + "num_tokens": 12606862.0, + "step": 1743 + }, + { + "epoch": 1.7777777777777777, + "grad_norm": 5.384610176086426, + "learning_rate": 1e-06, + "loss": 0.2824, + "mean_token_accuracy": 0.8926548957824707, + "num_tokens": 12614096.0, + "step": 1744 + }, + { + "epoch": 1.7787971457696228, + "grad_norm": 5.167020797729492, + "learning_rate": 1e-06, + "loss": 0.3176, + "mean_token_accuracy": 0.876054584980011, + "num_tokens": 12622668.0, + "step": 1745 + }, + { + "epoch": 1.7798165137614679, + "grad_norm": 5.2214813232421875, + "learning_rate": 1e-06, + "loss": 0.2734, + "mean_token_accuracy": 0.8902878165245056, + "num_tokens": 12630934.0, + "step": 1746 + }, + { + "epoch": 1.780835881753313, + "grad_norm": 5.939187049865723, + "learning_rate": 1e-06, + "loss": 0.3457, + "mean_token_accuracy": 0.8665916323661804, + "num_tokens": 12638097.0, + "step": 1747 + }, + { + "epoch": 1.781855249745158, + "grad_norm": 7.0170207023620605, + "learning_rate": 1e-06, + "loss": 0.3716, + "mean_token_accuracy": 0.8548344969749451, + "num_tokens": 12644199.0, + "step": 1748 + }, + { + "epoch": 1.782874617737003, + "grad_norm": 5.69582462310791, + "learning_rate": 1e-06, + "loss": 0.3304, + "mean_token_accuracy": 0.8711000680923462, + "num_tokens": 12651098.0, + "step": 1749 + }, + { + "epoch": 1.7838939857288483, + "grad_norm": 4.643102169036865, + "learning_rate": 1e-06, + "loss": 0.2777, + "mean_token_accuracy": 0.8911433219909668, + "num_tokens": 12658468.0, + "step": 1750 + }, + { + "epoch": 1.7849133537206932, + "grad_norm": 5.137328624725342, + "learning_rate": 1e-06, + "loss": 0.3228, + "mean_token_accuracy": 0.8746340870857239, + "num_tokens": 12666824.0, + "step": 1751 + }, + { + "epoch": 1.7859327217125383, + "grad_norm": 5.265320301055908, + "learning_rate": 1e-06, + "loss": 0.3053, + "mean_token_accuracy": 0.8777379989624023, + "num_tokens": 12675343.0, + "step": 1752 + }, + { + "epoch": 1.7869520897043834, + "grad_norm": 6.248562335968018, + "learning_rate": 1e-06, + "loss": 0.3639, + "mean_token_accuracy": 0.860029935836792, + "num_tokens": 12682000.0, + "step": 1753 + }, + { + "epoch": 1.7879714576962282, + "grad_norm": 5.902932167053223, + "learning_rate": 1e-06, + "loss": 0.2919, + "mean_token_accuracy": 0.8855603933334351, + "num_tokens": 12689125.0, + "step": 1754 + }, + { + "epoch": 1.7889908256880735, + "grad_norm": 4.538823127746582, + "learning_rate": 1e-06, + "loss": 0.2927, + "mean_token_accuracy": 0.8726692199707031, + "num_tokens": 12697508.0, + "step": 1755 + }, + { + "epoch": 1.7900101936799184, + "grad_norm": 5.676876544952393, + "learning_rate": 1e-06, + "loss": 0.3283, + "mean_token_accuracy": 0.8678065538406372, + "num_tokens": 12703796.0, + "step": 1756 + }, + { + "epoch": 1.7910295616717635, + "grad_norm": 4.702768325805664, + "learning_rate": 1e-06, + "loss": 0.2932, + "mean_token_accuracy": 0.8859347105026245, + "num_tokens": 12713555.0, + "step": 1757 + }, + { + "epoch": 1.7920489296636086, + "grad_norm": 5.5257134437561035, + "learning_rate": 1e-06, + "loss": 0.3061, + "mean_token_accuracy": 0.8744180202484131, + "num_tokens": 12720669.0, + "step": 1758 + }, + { + "epoch": 1.7930682976554535, + "grad_norm": 5.110534191131592, + "learning_rate": 1e-06, + "loss": 0.2999, + "mean_token_accuracy": 0.8732240796089172, + "num_tokens": 12727820.0, + "step": 1759 + }, + { + "epoch": 1.7940876656472988, + "grad_norm": 5.955296993255615, + "learning_rate": 1e-06, + "loss": 0.3333, + "mean_token_accuracy": 0.8715596795082092, + "num_tokens": 12734771.0, + "step": 1760 + }, + { + "epoch": 1.7951070336391437, + "grad_norm": 4.68795919418335, + "learning_rate": 1e-06, + "loss": 0.2662, + "mean_token_accuracy": 0.8966450691223145, + "num_tokens": 12743109.0, + "step": 1761 + }, + { + "epoch": 1.7961264016309888, + "grad_norm": 5.454711437225342, + "learning_rate": 1e-06, + "loss": 0.2958, + "mean_token_accuracy": 0.881873607635498, + "num_tokens": 12750367.0, + "step": 1762 + }, + { + "epoch": 1.797145769622834, + "grad_norm": 4.8620829582214355, + "learning_rate": 1e-06, + "loss": 0.273, + "mean_token_accuracy": 0.8838405609130859, + "num_tokens": 12759115.0, + "step": 1763 + }, + { + "epoch": 1.7981651376146788, + "grad_norm": 6.140048980712891, + "learning_rate": 1e-06, + "loss": 0.2969, + "mean_token_accuracy": 0.8781904578208923, + "num_tokens": 12765763.0, + "step": 1764 + }, + { + "epoch": 1.799184505606524, + "grad_norm": 6.207531929016113, + "learning_rate": 1e-06, + "loss": 0.3154, + "mean_token_accuracy": 0.8733164072036743, + "num_tokens": 12772993.0, + "step": 1765 + }, + { + "epoch": 1.800203873598369, + "grad_norm": 5.146313667297363, + "learning_rate": 1e-06, + "loss": 0.2768, + "mean_token_accuracy": 0.8911830186843872, + "num_tokens": 12780784.0, + "step": 1766 + }, + { + "epoch": 1.801223241590214, + "grad_norm": 5.319157123565674, + "learning_rate": 1e-06, + "loss": 0.3358, + "mean_token_accuracy": 0.86954665184021, + "num_tokens": 12788470.0, + "step": 1767 + }, + { + "epoch": 1.8022426095820592, + "grad_norm": 6.199775218963623, + "learning_rate": 1e-06, + "loss": 0.3349, + "mean_token_accuracy": 0.866553544998169, + "num_tokens": 12795007.0, + "step": 1768 + }, + { + "epoch": 1.803261977573904, + "grad_norm": 5.06267786026001, + "learning_rate": 1e-06, + "loss": 0.2852, + "mean_token_accuracy": 0.8883333206176758, + "num_tokens": 12803262.0, + "step": 1769 + }, + { + "epoch": 1.8042813455657494, + "grad_norm": 5.826524257659912, + "learning_rate": 1e-06, + "loss": 0.33, + "mean_token_accuracy": 0.8688356876373291, + "num_tokens": 12810564.0, + "step": 1770 + }, + { + "epoch": 1.8053007135575942, + "grad_norm": 7.028099536895752, + "learning_rate": 1e-06, + "loss": 0.3378, + "mean_token_accuracy": 0.8671998977661133, + "num_tokens": 12815509.0, + "step": 1771 + }, + { + "epoch": 1.8063200815494393, + "grad_norm": 5.445457458496094, + "learning_rate": 1e-06, + "loss": 0.3119, + "mean_token_accuracy": 0.8754478096961975, + "num_tokens": 12822642.0, + "step": 1772 + }, + { + "epoch": 1.8073394495412844, + "grad_norm": 6.70836067199707, + "learning_rate": 1e-06, + "loss": 0.3838, + "mean_token_accuracy": 0.8516464233398438, + "num_tokens": 12828530.0, + "step": 1773 + }, + { + "epoch": 1.8083588175331293, + "grad_norm": 5.61090087890625, + "learning_rate": 1e-06, + "loss": 0.3603, + "mean_token_accuracy": 0.8579515218734741, + "num_tokens": 12835390.0, + "step": 1774 + }, + { + "epoch": 1.8093781855249746, + "grad_norm": 5.963411808013916, + "learning_rate": 1e-06, + "loss": 0.3241, + "mean_token_accuracy": 0.8703550100326538, + "num_tokens": 12842001.0, + "step": 1775 + }, + { + "epoch": 1.8103975535168195, + "grad_norm": 4.88645601272583, + "learning_rate": 1e-06, + "loss": 0.2969, + "mean_token_accuracy": 0.881826639175415, + "num_tokens": 12850376.0, + "step": 1776 + }, + { + "epoch": 1.8114169215086646, + "grad_norm": 5.261552810668945, + "learning_rate": 1e-06, + "loss": 0.3269, + "mean_token_accuracy": 0.8726726174354553, + "num_tokens": 12856719.0, + "step": 1777 + }, + { + "epoch": 1.8124362895005097, + "grad_norm": 6.03155517578125, + "learning_rate": 1e-06, + "loss": 0.384, + "mean_token_accuracy": 0.8437178134918213, + "num_tokens": 12863346.0, + "step": 1778 + }, + { + "epoch": 1.8134556574923546, + "grad_norm": 5.693334102630615, + "learning_rate": 1e-06, + "loss": 0.3442, + "mean_token_accuracy": 0.8638437986373901, + "num_tokens": 12871722.0, + "step": 1779 + }, + { + "epoch": 1.8144750254842, + "grad_norm": 5.731080532073975, + "learning_rate": 1e-06, + "loss": 0.3469, + "mean_token_accuracy": 0.8627400398254395, + "num_tokens": 12878368.0, + "step": 1780 + }, + { + "epoch": 1.8154943934760448, + "grad_norm": 4.892282009124756, + "learning_rate": 1e-06, + "loss": 0.2795, + "mean_token_accuracy": 0.8845535516738892, + "num_tokens": 12885196.0, + "step": 1781 + }, + { + "epoch": 1.81651376146789, + "grad_norm": 5.3951945304870605, + "learning_rate": 1e-06, + "loss": 0.3238, + "mean_token_accuracy": 0.8730833530426025, + "num_tokens": 12894206.0, + "step": 1782 + }, + { + "epoch": 1.817533129459735, + "grad_norm": 5.3805341720581055, + "learning_rate": 1e-06, + "loss": 0.3158, + "mean_token_accuracy": 0.8742884993553162, + "num_tokens": 12901459.0, + "step": 1783 + }, + { + "epoch": 1.8185524974515799, + "grad_norm": 7.047651290893555, + "learning_rate": 1e-06, + "loss": 0.3502, + "mean_token_accuracy": 0.8553022742271423, + "num_tokens": 12906580.0, + "step": 1784 + }, + { + "epoch": 1.8195718654434252, + "grad_norm": 5.663541316986084, + "learning_rate": 1e-06, + "loss": 0.3288, + "mean_token_accuracy": 0.8668415546417236, + "num_tokens": 12914819.0, + "step": 1785 + }, + { + "epoch": 1.82059123343527, + "grad_norm": 5.576243877410889, + "learning_rate": 1e-06, + "loss": 0.3274, + "mean_token_accuracy": 0.8696330785751343, + "num_tokens": 12923086.0, + "step": 1786 + }, + { + "epoch": 1.8216106014271152, + "grad_norm": 4.482541561126709, + "learning_rate": 1e-06, + "loss": 0.2713, + "mean_token_accuracy": 0.8976508975028992, + "num_tokens": 12931773.0, + "step": 1787 + }, + { + "epoch": 1.8226299694189603, + "grad_norm": 5.149659156799316, + "learning_rate": 1e-06, + "loss": 0.3066, + "mean_token_accuracy": 0.8834894895553589, + "num_tokens": 12938788.0, + "step": 1788 + }, + { + "epoch": 1.8236493374108051, + "grad_norm": 6.784392356872559, + "learning_rate": 1e-06, + "loss": 0.3008, + "mean_token_accuracy": 0.8839720487594604, + "num_tokens": 12946716.0, + "step": 1789 + }, + { + "epoch": 1.8246687054026505, + "grad_norm": 6.238534927368164, + "learning_rate": 1e-06, + "loss": 0.3491, + "mean_token_accuracy": 0.8584759831428528, + "num_tokens": 12953778.0, + "step": 1790 + }, + { + "epoch": 1.8256880733944953, + "grad_norm": 5.276950359344482, + "learning_rate": 1e-06, + "loss": 0.287, + "mean_token_accuracy": 0.884618878364563, + "num_tokens": 12960061.0, + "step": 1791 + }, + { + "epoch": 1.8267074413863404, + "grad_norm": 5.917864799499512, + "learning_rate": 1e-06, + "loss": 0.3217, + "mean_token_accuracy": 0.8732515573501587, + "num_tokens": 12966608.0, + "step": 1792 + }, + { + "epoch": 1.8277268093781855, + "grad_norm": 5.792933464050293, + "learning_rate": 1e-06, + "loss": 0.3353, + "mean_token_accuracy": 0.8722871541976929, + "num_tokens": 12973674.0, + "step": 1793 + }, + { + "epoch": 1.8287461773700304, + "grad_norm": 4.932338237762451, + "learning_rate": 1e-06, + "loss": 0.3202, + "mean_token_accuracy": 0.8801292777061462, + "num_tokens": 12981738.0, + "step": 1794 + }, + { + "epoch": 1.8297655453618757, + "grad_norm": 6.709647178649902, + "learning_rate": 1e-06, + "loss": 0.3796, + "mean_token_accuracy": 0.8453797698020935, + "num_tokens": 12987585.0, + "step": 1795 + }, + { + "epoch": 1.8307849133537206, + "grad_norm": 5.906397342681885, + "learning_rate": 1e-06, + "loss": 0.3487, + "mean_token_accuracy": 0.8606617450714111, + "num_tokens": 12994898.0, + "step": 1796 + }, + { + "epoch": 1.8318042813455657, + "grad_norm": 5.688615798950195, + "learning_rate": 1e-06, + "loss": 0.3088, + "mean_token_accuracy": 0.8777914047241211, + "num_tokens": 13001354.0, + "step": 1797 + }, + { + "epoch": 1.8328236493374108, + "grad_norm": 4.629817485809326, + "learning_rate": 1e-06, + "loss": 0.2765, + "mean_token_accuracy": 0.8883563876152039, + "num_tokens": 13010187.0, + "step": 1798 + }, + { + "epoch": 1.8338430173292557, + "grad_norm": 5.129160404205322, + "learning_rate": 1e-06, + "loss": 0.3246, + "mean_token_accuracy": 0.8700505495071411, + "num_tokens": 13018081.0, + "step": 1799 + }, + { + "epoch": 1.834862385321101, + "grad_norm": 5.775502681732178, + "learning_rate": 1e-06, + "loss": 0.3433, + "mean_token_accuracy": 0.8674802780151367, + "num_tokens": 13024799.0, + "step": 1800 + }, + { + "epoch": 1.835881753312946, + "grad_norm": 5.379571437835693, + "learning_rate": 1e-06, + "loss": 0.3383, + "mean_token_accuracy": 0.8744577169418335, + "num_tokens": 13034348.0, + "step": 1801 + }, + { + "epoch": 1.836901121304791, + "grad_norm": 5.566597938537598, + "learning_rate": 1e-06, + "loss": 0.3372, + "mean_token_accuracy": 0.8735365271568298, + "num_tokens": 13042044.0, + "step": 1802 + }, + { + "epoch": 1.837920489296636, + "grad_norm": 4.924266338348389, + "learning_rate": 1e-06, + "loss": 0.288, + "mean_token_accuracy": 0.8864443302154541, + "num_tokens": 13049837.0, + "step": 1803 + }, + { + "epoch": 1.8389398572884812, + "grad_norm": 4.308236122131348, + "learning_rate": 1e-06, + "loss": 0.2815, + "mean_token_accuracy": 0.8832138776779175, + "num_tokens": 13058879.0, + "step": 1804 + }, + { + "epoch": 1.8399592252803263, + "grad_norm": 5.259842395782471, + "learning_rate": 1e-06, + "loss": 0.3125, + "mean_token_accuracy": 0.8779951930046082, + "num_tokens": 13066688.0, + "step": 1805 + }, + { + "epoch": 1.8409785932721712, + "grad_norm": 5.607459545135498, + "learning_rate": 1e-06, + "loss": 0.3357, + "mean_token_accuracy": 0.8718221187591553, + "num_tokens": 13073167.0, + "step": 1806 + }, + { + "epoch": 1.8419979612640163, + "grad_norm": 5.676422119140625, + "learning_rate": 1e-06, + "loss": 0.2997, + "mean_token_accuracy": 0.8776980638504028, + "num_tokens": 13080458.0, + "step": 1807 + }, + { + "epoch": 1.8430173292558614, + "grad_norm": 6.366734027862549, + "learning_rate": 1e-06, + "loss": 0.3462, + "mean_token_accuracy": 0.858220100402832, + "num_tokens": 13086830.0, + "step": 1808 + }, + { + "epoch": 1.8440366972477065, + "grad_norm": 5.100770950317383, + "learning_rate": 1e-06, + "loss": 0.2931, + "mean_token_accuracy": 0.8833786249160767, + "num_tokens": 13094692.0, + "step": 1809 + }, + { + "epoch": 1.8450560652395516, + "grad_norm": 6.3376007080078125, + "learning_rate": 1e-06, + "loss": 0.3751, + "mean_token_accuracy": 0.850662350654602, + "num_tokens": 13100914.0, + "step": 1810 + }, + { + "epoch": 1.8460754332313964, + "grad_norm": 4.5055646896362305, + "learning_rate": 1e-06, + "loss": 0.2662, + "mean_token_accuracy": 0.8978951573371887, + "num_tokens": 13109625.0, + "step": 1811 + }, + { + "epoch": 1.8470948012232415, + "grad_norm": 5.781276226043701, + "learning_rate": 1e-06, + "loss": 0.3468, + "mean_token_accuracy": 0.8607049584388733, + "num_tokens": 13116190.0, + "step": 1812 + }, + { + "epoch": 1.8481141692150866, + "grad_norm": 4.949795246124268, + "learning_rate": 1e-06, + "loss": 0.3141, + "mean_token_accuracy": 0.8711552619934082, + "num_tokens": 13123910.0, + "step": 1813 + }, + { + "epoch": 1.8491335372069317, + "grad_norm": 5.420562744140625, + "learning_rate": 1e-06, + "loss": 0.317, + "mean_token_accuracy": 0.8707503080368042, + "num_tokens": 13130692.0, + "step": 1814 + }, + { + "epoch": 1.8501529051987768, + "grad_norm": 5.026621341705322, + "learning_rate": 1e-06, + "loss": 0.2832, + "mean_token_accuracy": 0.8841001391410828, + "num_tokens": 13137525.0, + "step": 1815 + }, + { + "epoch": 1.8511722731906217, + "grad_norm": 6.480235576629639, + "learning_rate": 1e-06, + "loss": 0.3404, + "mean_token_accuracy": 0.8690832853317261, + "num_tokens": 13143436.0, + "step": 1816 + }, + { + "epoch": 1.8521916411824668, + "grad_norm": 5.9601521492004395, + "learning_rate": 1e-06, + "loss": 0.349, + "mean_token_accuracy": 0.8633898496627808, + "num_tokens": 13150215.0, + "step": 1817 + }, + { + "epoch": 1.853211009174312, + "grad_norm": 6.687257766723633, + "learning_rate": 1e-06, + "loss": 0.3305, + "mean_token_accuracy": 0.8583266139030457, + "num_tokens": 13157107.0, + "step": 1818 + }, + { + "epoch": 1.854230377166157, + "grad_norm": 4.824250221252441, + "learning_rate": 1e-06, + "loss": 0.2974, + "mean_token_accuracy": 0.8841132521629333, + "num_tokens": 13165786.0, + "step": 1819 + }, + { + "epoch": 1.8552497451580021, + "grad_norm": 5.853229999542236, + "learning_rate": 1e-06, + "loss": 0.3721, + "mean_token_accuracy": 0.8617588877677917, + "num_tokens": 13172930.0, + "step": 1820 + }, + { + "epoch": 1.856269113149847, + "grad_norm": 4.3974809646606445, + "learning_rate": 1e-06, + "loss": 0.2548, + "mean_token_accuracy": 0.8962990045547485, + "num_tokens": 13181475.0, + "step": 1821 + }, + { + "epoch": 1.8572884811416923, + "grad_norm": 6.370652675628662, + "learning_rate": 1e-06, + "loss": 0.3429, + "mean_token_accuracy": 0.8618543744087219, + "num_tokens": 13187515.0, + "step": 1822 + }, + { + "epoch": 1.8583078491335372, + "grad_norm": 5.544228553771973, + "learning_rate": 1e-06, + "loss": 0.3402, + "mean_token_accuracy": 0.8719897866249084, + "num_tokens": 13194960.0, + "step": 1823 + }, + { + "epoch": 1.8593272171253823, + "grad_norm": 6.008388519287109, + "learning_rate": 1e-06, + "loss": 0.3145, + "mean_token_accuracy": 0.8698028922080994, + "num_tokens": 13202032.0, + "step": 1824 + }, + { + "epoch": 1.8603465851172274, + "grad_norm": 4.873236179351807, + "learning_rate": 1e-06, + "loss": 0.3028, + "mean_token_accuracy": 0.8847393989562988, + "num_tokens": 13210649.0, + "step": 1825 + }, + { + "epoch": 1.8613659531090723, + "grad_norm": 5.510075569152832, + "learning_rate": 1e-06, + "loss": 0.3214, + "mean_token_accuracy": 0.876402735710144, + "num_tokens": 13218041.0, + "step": 1826 + }, + { + "epoch": 1.8623853211009176, + "grad_norm": 4.8013811111450195, + "learning_rate": 1e-06, + "loss": 0.2697, + "mean_token_accuracy": 0.8932369947433472, + "num_tokens": 13225755.0, + "step": 1827 + }, + { + "epoch": 1.8634046890927625, + "grad_norm": 5.332202911376953, + "learning_rate": 1e-06, + "loss": 0.2803, + "mean_token_accuracy": 0.8929609060287476, + "num_tokens": 13233561.0, + "step": 1828 + }, + { + "epoch": 1.8644240570846076, + "grad_norm": 5.74165678024292, + "learning_rate": 1e-06, + "loss": 0.3516, + "mean_token_accuracy": 0.8552602529525757, + "num_tokens": 13239832.0, + "step": 1829 + }, + { + "epoch": 1.8654434250764527, + "grad_norm": 5.895830154418945, + "learning_rate": 1e-06, + "loss": 0.3206, + "mean_token_accuracy": 0.8771979808807373, + "num_tokens": 13246517.0, + "step": 1830 + }, + { + "epoch": 1.8664627930682975, + "grad_norm": 5.336491107940674, + "learning_rate": 1e-06, + "loss": 0.3142, + "mean_token_accuracy": 0.8770743608474731, + "num_tokens": 13253797.0, + "step": 1831 + }, + { + "epoch": 1.8674821610601429, + "grad_norm": 6.684528350830078, + "learning_rate": 1e-06, + "loss": 0.3492, + "mean_token_accuracy": 0.8577903509140015, + "num_tokens": 13259758.0, + "step": 1832 + }, + { + "epoch": 1.8685015290519877, + "grad_norm": 6.736559867858887, + "learning_rate": 1e-06, + "loss": 0.3169, + "mean_token_accuracy": 0.8762653470039368, + "num_tokens": 13265847.0, + "step": 1833 + }, + { + "epoch": 1.8695208970438328, + "grad_norm": 4.994755268096924, + "learning_rate": 1e-06, + "loss": 0.2934, + "mean_token_accuracy": 0.8812415599822998, + "num_tokens": 13274190.0, + "step": 1834 + }, + { + "epoch": 1.870540265035678, + "grad_norm": 5.645715713500977, + "learning_rate": 1e-06, + "loss": 0.3074, + "mean_token_accuracy": 0.8752603530883789, + "num_tokens": 13280606.0, + "step": 1835 + }, + { + "epoch": 1.8715596330275228, + "grad_norm": 6.402475833892822, + "learning_rate": 1e-06, + "loss": 0.3419, + "mean_token_accuracy": 0.8661542534828186, + "num_tokens": 13287473.0, + "step": 1836 + }, + { + "epoch": 1.8725790010193681, + "grad_norm": 5.325406551361084, + "learning_rate": 1e-06, + "loss": 0.2951, + "mean_token_accuracy": 0.8794150352478027, + "num_tokens": 13295243.0, + "step": 1837 + }, + { + "epoch": 1.873598369011213, + "grad_norm": 5.593308448791504, + "learning_rate": 1e-06, + "loss": 0.3269, + "mean_token_accuracy": 0.876095175743103, + "num_tokens": 13302016.0, + "step": 1838 + }, + { + "epoch": 1.8746177370030581, + "grad_norm": 6.002068519592285, + "learning_rate": 1e-06, + "loss": 0.3196, + "mean_token_accuracy": 0.873113751411438, + "num_tokens": 13308889.0, + "step": 1839 + }, + { + "epoch": 1.8756371049949032, + "grad_norm": 6.207196235656738, + "learning_rate": 1e-06, + "loss": 0.3535, + "mean_token_accuracy": 0.8555775880813599, + "num_tokens": 13315542.0, + "step": 1840 + }, + { + "epoch": 1.876656472986748, + "grad_norm": 4.755486011505127, + "learning_rate": 1e-06, + "loss": 0.2747, + "mean_token_accuracy": 0.8901534080505371, + "num_tokens": 13323383.0, + "step": 1841 + }, + { + "epoch": 1.8776758409785934, + "grad_norm": 5.736021995544434, + "learning_rate": 1e-06, + "loss": 0.3629, + "mean_token_accuracy": 0.8601976633071899, + "num_tokens": 13329601.0, + "step": 1842 + }, + { + "epoch": 1.8786952089704383, + "grad_norm": 5.392532825469971, + "learning_rate": 1e-06, + "loss": 0.322, + "mean_token_accuracy": 0.8735746145248413, + "num_tokens": 13336692.0, + "step": 1843 + }, + { + "epoch": 1.8797145769622834, + "grad_norm": 5.750321388244629, + "learning_rate": 1e-06, + "loss": 0.3213, + "mean_token_accuracy": 0.8627414107322693, + "num_tokens": 13343341.0, + "step": 1844 + }, + { + "epoch": 1.8807339449541285, + "grad_norm": 5.433614253997803, + "learning_rate": 1e-06, + "loss": 0.2995, + "mean_token_accuracy": 0.8849574327468872, + "num_tokens": 13350025.0, + "step": 1845 + }, + { + "epoch": 1.8817533129459734, + "grad_norm": 6.240256309509277, + "learning_rate": 1e-06, + "loss": 0.3058, + "mean_token_accuracy": 0.8798707723617554, + "num_tokens": 13356310.0, + "step": 1846 + }, + { + "epoch": 1.8827726809378187, + "grad_norm": 4.441460132598877, + "learning_rate": 1e-06, + "loss": 0.2889, + "mean_token_accuracy": 0.8893877267837524, + "num_tokens": 13365000.0, + "step": 1847 + }, + { + "epoch": 1.8837920489296636, + "grad_norm": 6.006625175476074, + "learning_rate": 1e-06, + "loss": 0.3182, + "mean_token_accuracy": 0.8780320286750793, + "num_tokens": 13371891.0, + "step": 1848 + }, + { + "epoch": 1.8848114169215087, + "grad_norm": 5.326073169708252, + "learning_rate": 1e-06, + "loss": 0.2994, + "mean_token_accuracy": 0.8852208256721497, + "num_tokens": 13379967.0, + "step": 1849 + }, + { + "epoch": 1.8858307849133538, + "grad_norm": 5.640017509460449, + "learning_rate": 1e-06, + "loss": 0.3509, + "mean_token_accuracy": 0.8592500686645508, + "num_tokens": 13386963.0, + "step": 1850 + }, + { + "epoch": 1.8868501529051986, + "grad_norm": 5.606468677520752, + "learning_rate": 1e-06, + "loss": 0.3057, + "mean_token_accuracy": 0.8807944059371948, + "num_tokens": 13394669.0, + "step": 1851 + }, + { + "epoch": 1.887869520897044, + "grad_norm": 6.065982818603516, + "learning_rate": 1e-06, + "loss": 0.3734, + "mean_token_accuracy": 0.8537954092025757, + "num_tokens": 13400998.0, + "step": 1852 + }, + { + "epoch": 1.8888888888888888, + "grad_norm": 5.216724395751953, + "learning_rate": 1e-06, + "loss": 0.3223, + "mean_token_accuracy": 0.8704068064689636, + "num_tokens": 13408569.0, + "step": 1853 + }, + { + "epoch": 1.889908256880734, + "grad_norm": 5.510074615478516, + "learning_rate": 1e-06, + "loss": 0.3225, + "mean_token_accuracy": 0.872739315032959, + "num_tokens": 13415180.0, + "step": 1854 + }, + { + "epoch": 1.890927624872579, + "grad_norm": 6.225845813751221, + "learning_rate": 1e-06, + "loss": 0.3628, + "mean_token_accuracy": 0.8667094707489014, + "num_tokens": 13420956.0, + "step": 1855 + }, + { + "epoch": 1.891946992864424, + "grad_norm": 4.401638984680176, + "learning_rate": 1e-06, + "loss": 0.3101, + "mean_token_accuracy": 0.8689122200012207, + "num_tokens": 13429288.0, + "step": 1856 + }, + { + "epoch": 1.8929663608562692, + "grad_norm": 4.500893592834473, + "learning_rate": 1e-06, + "loss": 0.3088, + "mean_token_accuracy": 0.8753281235694885, + "num_tokens": 13438573.0, + "step": 1857 + }, + { + "epoch": 1.8939857288481141, + "grad_norm": 4.655992031097412, + "learning_rate": 1e-06, + "loss": 0.3242, + "mean_token_accuracy": 0.8742519617080688, + "num_tokens": 13447378.0, + "step": 1858 + }, + { + "epoch": 1.8950050968399592, + "grad_norm": 5.0165300369262695, + "learning_rate": 1e-06, + "loss": 0.2643, + "mean_token_accuracy": 0.8820810317993164, + "num_tokens": 13454948.0, + "step": 1859 + }, + { + "epoch": 1.8960244648318043, + "grad_norm": 5.494775295257568, + "learning_rate": 1e-06, + "loss": 0.3286, + "mean_token_accuracy": 0.8667450547218323, + "num_tokens": 13461872.0, + "step": 1860 + }, + { + "epoch": 1.8970438328236492, + "grad_norm": 5.0786237716674805, + "learning_rate": 1e-06, + "loss": 0.313, + "mean_token_accuracy": 0.8764716982841492, + "num_tokens": 13470060.0, + "step": 1861 + }, + { + "epoch": 1.8980632008154945, + "grad_norm": 5.586287498474121, + "learning_rate": 1e-06, + "loss": 0.3041, + "mean_token_accuracy": 0.8937321901321411, + "num_tokens": 13477388.0, + "step": 1862 + }, + { + "epoch": 1.8990825688073394, + "grad_norm": 6.048027038574219, + "learning_rate": 1e-06, + "loss": 0.3652, + "mean_token_accuracy": 0.8603817224502563, + "num_tokens": 13483456.0, + "step": 1863 + }, + { + "epoch": 1.9001019367991845, + "grad_norm": 5.646549701690674, + "learning_rate": 1e-06, + "loss": 0.2847, + "mean_token_accuracy": 0.8880375623703003, + "num_tokens": 13489932.0, + "step": 1864 + }, + { + "epoch": 1.9011213047910296, + "grad_norm": 5.600374221801758, + "learning_rate": 1e-06, + "loss": 0.3434, + "mean_token_accuracy": 0.8619102239608765, + "num_tokens": 13496786.0, + "step": 1865 + }, + { + "epoch": 1.9021406727828745, + "grad_norm": 5.790276527404785, + "learning_rate": 1e-06, + "loss": 0.3427, + "mean_token_accuracy": 0.8683071136474609, + "num_tokens": 13503863.0, + "step": 1866 + }, + { + "epoch": 1.9031600407747198, + "grad_norm": 5.744128704071045, + "learning_rate": 1e-06, + "loss": 0.3347, + "mean_token_accuracy": 0.8681007623672485, + "num_tokens": 13510024.0, + "step": 1867 + }, + { + "epoch": 1.9041794087665647, + "grad_norm": 5.457881927490234, + "learning_rate": 1e-06, + "loss": 0.2824, + "mean_token_accuracy": 0.8873980045318604, + "num_tokens": 13517009.0, + "step": 1868 + }, + { + "epoch": 1.9051987767584098, + "grad_norm": 5.751425266265869, + "learning_rate": 1e-06, + "loss": 0.3313, + "mean_token_accuracy": 0.8667054176330566, + "num_tokens": 13523767.0, + "step": 1869 + }, + { + "epoch": 1.9062181447502549, + "grad_norm": 6.1151299476623535, + "learning_rate": 1e-06, + "loss": 0.3097, + "mean_token_accuracy": 0.8735657930374146, + "num_tokens": 13530357.0, + "step": 1870 + }, + { + "epoch": 1.9072375127420997, + "grad_norm": 5.2769036293029785, + "learning_rate": 1e-06, + "loss": 0.303, + "mean_token_accuracy": 0.8823283910751343, + "num_tokens": 13538330.0, + "step": 1871 + }, + { + "epoch": 1.908256880733945, + "grad_norm": 6.263132095336914, + "learning_rate": 1e-06, + "loss": 0.3131, + "mean_token_accuracy": 0.8731421232223511, + "num_tokens": 13544230.0, + "step": 1872 + }, + { + "epoch": 1.90927624872579, + "grad_norm": 5.496922492980957, + "learning_rate": 1e-06, + "loss": 0.3141, + "mean_token_accuracy": 0.8701661825180054, + "num_tokens": 13550675.0, + "step": 1873 + }, + { + "epoch": 1.910295616717635, + "grad_norm": 5.973094940185547, + "learning_rate": 1e-06, + "loss": 0.3279, + "mean_token_accuracy": 0.871132493019104, + "num_tokens": 13557751.0, + "step": 1874 + }, + { + "epoch": 1.9113149847094801, + "grad_norm": 6.0180840492248535, + "learning_rate": 1e-06, + "loss": 0.3467, + "mean_token_accuracy": 0.8745821118354797, + "num_tokens": 13565296.0, + "step": 1875 + }, + { + "epoch": 1.912334352701325, + "grad_norm": 5.1452155113220215, + "learning_rate": 1e-06, + "loss": 0.3294, + "mean_token_accuracy": 0.8702189326286316, + "num_tokens": 13573759.0, + "step": 1876 + }, + { + "epoch": 1.9133537206931703, + "grad_norm": 5.384848594665527, + "learning_rate": 1e-06, + "loss": 0.3413, + "mean_token_accuracy": 0.8614591956138611, + "num_tokens": 13580638.0, + "step": 1877 + }, + { + "epoch": 1.9143730886850152, + "grad_norm": 4.4838643074035645, + "learning_rate": 1e-06, + "loss": 0.2606, + "mean_token_accuracy": 0.8883817195892334, + "num_tokens": 13589288.0, + "step": 1878 + }, + { + "epoch": 1.9153924566768603, + "grad_norm": 4.459898948669434, + "learning_rate": 1e-06, + "loss": 0.2676, + "mean_token_accuracy": 0.8941344618797302, + "num_tokens": 13597638.0, + "step": 1879 + }, + { + "epoch": 1.9164118246687054, + "grad_norm": 5.410155773162842, + "learning_rate": 1e-06, + "loss": 0.3065, + "mean_token_accuracy": 0.8788959980010986, + "num_tokens": 13604527.0, + "step": 1880 + }, + { + "epoch": 1.9174311926605505, + "grad_norm": 4.80377721786499, + "learning_rate": 1e-06, + "loss": 0.317, + "mean_token_accuracy": 0.8727308511734009, + "num_tokens": 13613444.0, + "step": 1881 + }, + { + "epoch": 1.9184505606523956, + "grad_norm": 4.696110725402832, + "learning_rate": 1e-06, + "loss": 0.2837, + "mean_token_accuracy": 0.8882883787155151, + "num_tokens": 13622770.0, + "step": 1882 + }, + { + "epoch": 1.9194699286442405, + "grad_norm": 5.143455982208252, + "learning_rate": 1e-06, + "loss": 0.2843, + "mean_token_accuracy": 0.8850981593132019, + "num_tokens": 13630548.0, + "step": 1883 + }, + { + "epoch": 1.9204892966360856, + "grad_norm": 5.234915733337402, + "learning_rate": 1e-06, + "loss": 0.3199, + "mean_token_accuracy": 0.8655498027801514, + "num_tokens": 13638563.0, + "step": 1884 + }, + { + "epoch": 1.9215086646279307, + "grad_norm": 6.458122730255127, + "learning_rate": 1e-06, + "loss": 0.3525, + "mean_token_accuracy": 0.8594453930854797, + "num_tokens": 13645472.0, + "step": 1885 + }, + { + "epoch": 1.9225280326197758, + "grad_norm": 6.209213733673096, + "learning_rate": 1e-06, + "loss": 0.3149, + "mean_token_accuracy": 0.8773858547210693, + "num_tokens": 13651404.0, + "step": 1886 + }, + { + "epoch": 1.9235474006116209, + "grad_norm": 5.333641529083252, + "learning_rate": 1e-06, + "loss": 0.32, + "mean_token_accuracy": 0.8773972988128662, + "num_tokens": 13658041.0, + "step": 1887 + }, + { + "epoch": 1.9245667686034658, + "grad_norm": 5.1544013023376465, + "learning_rate": 1e-06, + "loss": 0.2773, + "mean_token_accuracy": 0.897637128829956, + "num_tokens": 13666344.0, + "step": 1888 + }, + { + "epoch": 1.9255861365953109, + "grad_norm": 5.681913375854492, + "learning_rate": 1e-06, + "loss": 0.3167, + "mean_token_accuracy": 0.8704078197479248, + "num_tokens": 13673583.0, + "step": 1889 + }, + { + "epoch": 1.926605504587156, + "grad_norm": 5.437119960784912, + "learning_rate": 1e-06, + "loss": 0.3323, + "mean_token_accuracy": 0.8749741911888123, + "num_tokens": 13680170.0, + "step": 1890 + }, + { + "epoch": 1.927624872579001, + "grad_norm": 5.13888692855835, + "learning_rate": 1e-06, + "loss": 0.3398, + "mean_token_accuracy": 0.8666523694992065, + "num_tokens": 13688421.0, + "step": 1891 + }, + { + "epoch": 1.9286442405708462, + "grad_norm": 6.75938081741333, + "learning_rate": 1e-06, + "loss": 0.3516, + "mean_token_accuracy": 0.8535584211349487, + "num_tokens": 13694494.0, + "step": 1892 + }, + { + "epoch": 1.929663608562691, + "grad_norm": 5.392444610595703, + "learning_rate": 1e-06, + "loss": 0.3278, + "mean_token_accuracy": 0.869667649269104, + "num_tokens": 13701774.0, + "step": 1893 + }, + { + "epoch": 1.9306829765545361, + "grad_norm": 5.634164333343506, + "learning_rate": 1e-06, + "loss": 0.3483, + "mean_token_accuracy": 0.8651007413864136, + "num_tokens": 13710126.0, + "step": 1894 + }, + { + "epoch": 1.9317023445463812, + "grad_norm": 5.962131977081299, + "learning_rate": 1e-06, + "loss": 0.3218, + "mean_token_accuracy": 0.8694875240325928, + "num_tokens": 13716115.0, + "step": 1895 + }, + { + "epoch": 1.9327217125382263, + "grad_norm": 5.609973907470703, + "learning_rate": 1e-06, + "loss": 0.3205, + "mean_token_accuracy": 0.8659988641738892, + "num_tokens": 13723412.0, + "step": 1896 + }, + { + "epoch": 1.9337410805300714, + "grad_norm": 4.6243062019348145, + "learning_rate": 1e-06, + "loss": 0.2973, + "mean_token_accuracy": 0.8869944214820862, + "num_tokens": 13731460.0, + "step": 1897 + }, + { + "epoch": 1.9347604485219163, + "grad_norm": 5.284345626831055, + "learning_rate": 1e-06, + "loss": 0.3269, + "mean_token_accuracy": 0.8767040967941284, + "num_tokens": 13738660.0, + "step": 1898 + }, + { + "epoch": 1.9357798165137616, + "grad_norm": 5.469335556030273, + "learning_rate": 1e-06, + "loss": 0.3735, + "mean_token_accuracy": 0.8503420352935791, + "num_tokens": 13746423.0, + "step": 1899 + }, + { + "epoch": 1.9367991845056065, + "grad_norm": 4.687780380249023, + "learning_rate": 1e-06, + "loss": 0.3109, + "mean_token_accuracy": 0.8763988614082336, + "num_tokens": 13754536.0, + "step": 1900 + }, + { + "epoch": 1.9378185524974516, + "grad_norm": 5.242497444152832, + "learning_rate": 1e-06, + "loss": 0.3073, + "mean_token_accuracy": 0.8762999773025513, + "num_tokens": 13762437.0, + "step": 1901 + }, + { + "epoch": 1.9388379204892967, + "grad_norm": 5.508381366729736, + "learning_rate": 1e-06, + "loss": 0.3101, + "mean_token_accuracy": 0.8767776489257812, + "num_tokens": 13769129.0, + "step": 1902 + }, + { + "epoch": 1.9398572884811416, + "grad_norm": 5.4623918533325195, + "learning_rate": 1e-06, + "loss": 0.3517, + "mean_token_accuracy": 0.8622088432312012, + "num_tokens": 13776580.0, + "step": 1903 + }, + { + "epoch": 1.940876656472987, + "grad_norm": 5.770425319671631, + "learning_rate": 1e-06, + "loss": 0.3605, + "mean_token_accuracy": 0.8526716828346252, + "num_tokens": 13784222.0, + "step": 1904 + }, + { + "epoch": 1.9418960244648318, + "grad_norm": 5.185537338256836, + "learning_rate": 1e-06, + "loss": 0.2974, + "mean_token_accuracy": 0.8819513320922852, + "num_tokens": 13791124.0, + "step": 1905 + }, + { + "epoch": 1.9429153924566769, + "grad_norm": 5.451938152313232, + "learning_rate": 1e-06, + "loss": 0.3282, + "mean_token_accuracy": 0.8636451959609985, + "num_tokens": 13798182.0, + "step": 1906 + }, + { + "epoch": 1.943934760448522, + "grad_norm": 6.57783317565918, + "learning_rate": 1e-06, + "loss": 0.3429, + "mean_token_accuracy": 0.8642681241035461, + "num_tokens": 13803944.0, + "step": 1907 + }, + { + "epoch": 1.9449541284403669, + "grad_norm": 5.016219615936279, + "learning_rate": 1e-06, + "loss": 0.2872, + "mean_token_accuracy": 0.8893285989761353, + "num_tokens": 13811509.0, + "step": 1908 + }, + { + "epoch": 1.9459734964322122, + "grad_norm": 6.348482608795166, + "learning_rate": 1e-06, + "loss": 0.3299, + "mean_token_accuracy": 0.8692911863327026, + "num_tokens": 13816983.0, + "step": 1909 + }, + { + "epoch": 1.946992864424057, + "grad_norm": 5.602998733520508, + "learning_rate": 1e-06, + "loss": 0.2989, + "mean_token_accuracy": 0.881986677646637, + "num_tokens": 13824471.0, + "step": 1910 + }, + { + "epoch": 1.9480122324159022, + "grad_norm": 6.186687469482422, + "learning_rate": 1e-06, + "loss": 0.3291, + "mean_token_accuracy": 0.8601089119911194, + "num_tokens": 13831393.0, + "step": 1911 + }, + { + "epoch": 1.9490316004077473, + "grad_norm": 5.265104293823242, + "learning_rate": 1e-06, + "loss": 0.2819, + "mean_token_accuracy": 0.8880878686904907, + "num_tokens": 13838641.0, + "step": 1912 + }, + { + "epoch": 1.9500509683995921, + "grad_norm": 5.378809452056885, + "learning_rate": 1e-06, + "loss": 0.3408, + "mean_token_accuracy": 0.8657407760620117, + "num_tokens": 13846648.0, + "step": 1913 + }, + { + "epoch": 1.9510703363914375, + "grad_norm": 5.620635032653809, + "learning_rate": 1e-06, + "loss": 0.3233, + "mean_token_accuracy": 0.8706358075141907, + "num_tokens": 13853808.0, + "step": 1914 + }, + { + "epoch": 1.9520897043832823, + "grad_norm": 5.957886219024658, + "learning_rate": 1e-06, + "loss": 0.334, + "mean_token_accuracy": 0.8635218143463135, + "num_tokens": 13860705.0, + "step": 1915 + }, + { + "epoch": 1.9531090723751274, + "grad_norm": 6.112631797790527, + "learning_rate": 1e-06, + "loss": 0.3673, + "mean_token_accuracy": 0.8566216230392456, + "num_tokens": 13867472.0, + "step": 1916 + }, + { + "epoch": 1.9541284403669725, + "grad_norm": 5.579799175262451, + "learning_rate": 1e-06, + "loss": 0.326, + "mean_token_accuracy": 0.8670506477355957, + "num_tokens": 13875719.0, + "step": 1917 + }, + { + "epoch": 1.9551478083588174, + "grad_norm": 4.862008571624756, + "learning_rate": 1e-06, + "loss": 0.295, + "mean_token_accuracy": 0.8754022121429443, + "num_tokens": 13883975.0, + "step": 1918 + }, + { + "epoch": 1.9561671763506627, + "grad_norm": 6.821905612945557, + "learning_rate": 1e-06, + "loss": 0.3929, + "mean_token_accuracy": 0.8445662260055542, + "num_tokens": 13889571.0, + "step": 1919 + }, + { + "epoch": 1.9571865443425076, + "grad_norm": 5.289796352386475, + "learning_rate": 1e-06, + "loss": 0.3435, + "mean_token_accuracy": 0.8641958832740784, + "num_tokens": 13898189.0, + "step": 1920 + }, + { + "epoch": 1.9582059123343527, + "grad_norm": 5.393865585327148, + "learning_rate": 1e-06, + "loss": 0.2924, + "mean_token_accuracy": 0.8836370706558228, + "num_tokens": 13905385.0, + "step": 1921 + }, + { + "epoch": 1.9592252803261978, + "grad_norm": 5.3781514167785645, + "learning_rate": 1e-06, + "loss": 0.3443, + "mean_token_accuracy": 0.8632959127426147, + "num_tokens": 13911852.0, + "step": 1922 + }, + { + "epoch": 1.9602446483180427, + "grad_norm": 5.6426520347595215, + "learning_rate": 1e-06, + "loss": 0.2881, + "mean_token_accuracy": 0.8870062828063965, + "num_tokens": 13918762.0, + "step": 1923 + }, + { + "epoch": 1.961264016309888, + "grad_norm": 5.755259990692139, + "learning_rate": 1e-06, + "loss": 0.3455, + "mean_token_accuracy": 0.8650844097137451, + "num_tokens": 13925819.0, + "step": 1924 + }, + { + "epoch": 1.9622833843017329, + "grad_norm": 5.435282230377197, + "learning_rate": 1e-06, + "loss": 0.3463, + "mean_token_accuracy": 0.8633766770362854, + "num_tokens": 13933673.0, + "step": 1925 + }, + { + "epoch": 1.963302752293578, + "grad_norm": 4.551942825317383, + "learning_rate": 1e-06, + "loss": 0.3029, + "mean_token_accuracy": 0.8780261278152466, + "num_tokens": 13941406.0, + "step": 1926 + }, + { + "epoch": 1.964322120285423, + "grad_norm": 5.808569431304932, + "learning_rate": 1e-06, + "loss": 0.3097, + "mean_token_accuracy": 0.8838866353034973, + "num_tokens": 13947235.0, + "step": 1927 + }, + { + "epoch": 1.965341488277268, + "grad_norm": 5.91684627532959, + "learning_rate": 1e-06, + "loss": 0.3149, + "mean_token_accuracy": 0.8689714074134827, + "num_tokens": 13953776.0, + "step": 1928 + }, + { + "epoch": 1.9663608562691133, + "grad_norm": 5.612284183502197, + "learning_rate": 1e-06, + "loss": 0.3245, + "mean_token_accuracy": 0.868607223033905, + "num_tokens": 13961174.0, + "step": 1929 + }, + { + "epoch": 1.9673802242609582, + "grad_norm": 5.067611217498779, + "learning_rate": 1e-06, + "loss": 0.3402, + "mean_token_accuracy": 0.8669734001159668, + "num_tokens": 13969745.0, + "step": 1930 + }, + { + "epoch": 1.9683995922528033, + "grad_norm": 4.432025909423828, + "learning_rate": 1e-06, + "loss": 0.2729, + "mean_token_accuracy": 0.8919516801834106, + "num_tokens": 13978097.0, + "step": 1931 + }, + { + "epoch": 1.9694189602446484, + "grad_norm": 6.11106014251709, + "learning_rate": 1e-06, + "loss": 0.3307, + "mean_token_accuracy": 0.8672759532928467, + "num_tokens": 13983747.0, + "step": 1932 + }, + { + "epoch": 1.9704383282364932, + "grad_norm": 4.8504180908203125, + "learning_rate": 1e-06, + "loss": 0.3066, + "mean_token_accuracy": 0.8800613284111023, + "num_tokens": 13992458.0, + "step": 1933 + }, + { + "epoch": 1.9714576962283386, + "grad_norm": 4.7586212158203125, + "learning_rate": 1e-06, + "loss": 0.2846, + "mean_token_accuracy": 0.8820195198059082, + "num_tokens": 14000381.0, + "step": 1934 + }, + { + "epoch": 1.9724770642201834, + "grad_norm": 5.280855178833008, + "learning_rate": 1e-06, + "loss": 0.3189, + "mean_token_accuracy": 0.8773078918457031, + "num_tokens": 14007729.0, + "step": 1935 + }, + { + "epoch": 1.9734964322120285, + "grad_norm": 5.421816349029541, + "learning_rate": 1e-06, + "loss": 0.3082, + "mean_token_accuracy": 0.876941978931427, + "num_tokens": 14015434.0, + "step": 1936 + }, + { + "epoch": 1.9745158002038736, + "grad_norm": 4.949092388153076, + "learning_rate": 1e-06, + "loss": 0.2977, + "mean_token_accuracy": 0.8797652721405029, + "num_tokens": 14023192.0, + "step": 1937 + }, + { + "epoch": 1.9755351681957185, + "grad_norm": 5.960864067077637, + "learning_rate": 1e-06, + "loss": 0.3644, + "mean_token_accuracy": 0.8580238819122314, + "num_tokens": 14030453.0, + "step": 1938 + }, + { + "epoch": 1.9765545361875638, + "grad_norm": 5.306753635406494, + "learning_rate": 1e-06, + "loss": 0.3152, + "mean_token_accuracy": 0.8756694793701172, + "num_tokens": 14038720.0, + "step": 1939 + }, + { + "epoch": 1.9775739041794087, + "grad_norm": 4.490878105163574, + "learning_rate": 1e-06, + "loss": 0.2565, + "mean_token_accuracy": 0.8951671123504639, + "num_tokens": 14046648.0, + "step": 1940 + }, + { + "epoch": 1.9785932721712538, + "grad_norm": 5.744395732879639, + "learning_rate": 1e-06, + "loss": 0.3343, + "mean_token_accuracy": 0.8685741424560547, + "num_tokens": 14053084.0, + "step": 1941 + }, + { + "epoch": 1.979612640163099, + "grad_norm": 5.042215824127197, + "learning_rate": 1e-06, + "loss": 0.3061, + "mean_token_accuracy": 0.8817275762557983, + "num_tokens": 14061268.0, + "step": 1942 + }, + { + "epoch": 1.9806320081549438, + "grad_norm": 4.9874587059021, + "learning_rate": 1e-06, + "loss": 0.2836, + "mean_token_accuracy": 0.8894073963165283, + "num_tokens": 14068442.0, + "step": 1943 + }, + { + "epoch": 1.981651376146789, + "grad_norm": 5.447760581970215, + "learning_rate": 1e-06, + "loss": 0.3201, + "mean_token_accuracy": 0.8742286562919617, + "num_tokens": 14076125.0, + "step": 1944 + }, + { + "epoch": 1.982670744138634, + "grad_norm": 5.494638442993164, + "learning_rate": 1e-06, + "loss": 0.3463, + "mean_token_accuracy": 0.8655694723129272, + "num_tokens": 14083553.0, + "step": 1945 + }, + { + "epoch": 1.983690112130479, + "grad_norm": 5.815045356750488, + "learning_rate": 1e-06, + "loss": 0.3141, + "mean_token_accuracy": 0.869587779045105, + "num_tokens": 14089842.0, + "step": 1946 + }, + { + "epoch": 1.9847094801223242, + "grad_norm": 4.741557598114014, + "learning_rate": 1e-06, + "loss": 0.2928, + "mean_token_accuracy": 0.8860731720924377, + "num_tokens": 14097994.0, + "step": 1947 + }, + { + "epoch": 1.985728848114169, + "grad_norm": 4.5859527587890625, + "learning_rate": 1e-06, + "loss": 0.286, + "mean_token_accuracy": 0.8851276636123657, + "num_tokens": 14106396.0, + "step": 1948 + }, + { + "epoch": 1.9867482161060144, + "grad_norm": 6.422934055328369, + "learning_rate": 1e-06, + "loss": 0.3539, + "mean_token_accuracy": 0.8640207648277283, + "num_tokens": 14112785.0, + "step": 1949 + }, + { + "epoch": 1.9877675840978593, + "grad_norm": 5.655293941497803, + "learning_rate": 1e-06, + "loss": 0.3207, + "mean_token_accuracy": 0.8743720054626465, + "num_tokens": 14120398.0, + "step": 1950 + }, + { + "epoch": 1.9887869520897044, + "grad_norm": 6.411019802093506, + "learning_rate": 1e-06, + "loss": 0.2988, + "mean_token_accuracy": 0.8837896585464478, + "num_tokens": 14128388.0, + "step": 1951 + }, + { + "epoch": 1.9898063200815495, + "grad_norm": 5.964116096496582, + "learning_rate": 1e-06, + "loss": 0.3194, + "mean_token_accuracy": 0.8718749284744263, + "num_tokens": 14135396.0, + "step": 1952 + }, + { + "epoch": 1.9908256880733946, + "grad_norm": 5.702261924743652, + "learning_rate": 1e-06, + "loss": 0.3452, + "mean_token_accuracy": 0.8555079102516174, + "num_tokens": 14141672.0, + "step": 1953 + }, + { + "epoch": 1.9918450560652396, + "grad_norm": 4.808823585510254, + "learning_rate": 1e-06, + "loss": 0.3283, + "mean_token_accuracy": 0.8714936971664429, + "num_tokens": 14151504.0, + "step": 1954 + }, + { + "epoch": 1.9928644240570845, + "grad_norm": 5.689114570617676, + "learning_rate": 1e-06, + "loss": 0.3468, + "mean_token_accuracy": 0.8634822368621826, + "num_tokens": 14158355.0, + "step": 1955 + }, + { + "epoch": 1.9938837920489296, + "grad_norm": 4.6639628410339355, + "learning_rate": 1e-06, + "loss": 0.3144, + "mean_token_accuracy": 0.8738683462142944, + "num_tokens": 14166722.0, + "step": 1956 + }, + { + "epoch": 1.9949031600407747, + "grad_norm": 5.481828212738037, + "learning_rate": 1e-06, + "loss": 0.347, + "mean_token_accuracy": 0.8645076751708984, + "num_tokens": 14174261.0, + "step": 1957 + }, + { + "epoch": 1.9959225280326198, + "grad_norm": 5.878755569458008, + "learning_rate": 1e-06, + "loss": 0.3611, + "mean_token_accuracy": 0.855923593044281, + "num_tokens": 14180670.0, + "step": 1958 + }, + { + "epoch": 1.996941896024465, + "grad_norm": 5.106919765472412, + "learning_rate": 1e-06, + "loss": 0.2993, + "mean_token_accuracy": 0.8851836919784546, + "num_tokens": 14188706.0, + "step": 1959 + }, + { + "epoch": 1.9979612640163098, + "grad_norm": 4.610098361968994, + "learning_rate": 1e-06, + "loss": 0.2874, + "mean_token_accuracy": 0.8931455612182617, + "num_tokens": 14196253.0, + "step": 1960 + }, + { + "epoch": 1.998980632008155, + "grad_norm": 4.565979957580566, + "learning_rate": 1e-06, + "loss": 0.2943, + "mean_token_accuracy": 0.8854101896286011, + "num_tokens": 14206646.0, + "step": 1961 + }, + { + "epoch": 2.0, + "grad_norm": 5.152297019958496, + "learning_rate": 1e-06, + "loss": 0.3256, + "mean_token_accuracy": 0.8729977607727051, + "num_tokens": 14215368.0, + "step": 1962 + }, + { + "epoch": 2.0, + "step": 1962, + "total_flos": 6.401115957451817e+17, + "train_loss": 0.3628057053238609, + "train_runtime": 1540.8029, + "train_samples_per_second": 20.374, + "train_steps_per_second": 1.273 + } + ], + "logging_steps": 1, + "max_steps": 1962, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 981, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 6.401115957451817e+17, + "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..213b76a --- /dev/null +++ b/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0838c83e3127bcd9f02ce3350b7b8bbaf30fdef825d3e82fd9a69fb08b1ce13 +size 13329