[2026-04-16 10:31:52,083] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:135] baseline 0.000GB () [2026-04-16 10:31:52,086] [INFO] [axolotl.cli.config.load_cfg:341] [PID:135] config: { "activation_offloading": false, "axolotl_config_path": "./config.yaml", "base_model": "Qwen/Qwen2.5-7B", "base_model_config": "Qwen/Qwen2.5-7B", "batch_size": 16, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_80", "fp8": false, "n_gpu": 1, "n_node": 1, "tf32": true }, "context_parallel_size": 1, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_num_proc": 96, "dataset_prepared_path": "./prepared_data", "datasets": [ { "path": "Bossssss/medqa-dpo-idk", "split": "train", "type": "passthrough" } ], "ddp": false, "device": "cuda:0", "dion_rank_fraction": 1.0, "dion_rank_multiple_of": 1, "eaft_alpha": 1.0, "eaft_k": 20, "env_capabilities": { "torch_version": "2.9.1" }, "eval_batch_size": 2, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_steps": 100, "eval_strategy": "steps", "eval_table_size": 0, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "generate_samples": false, "generation_do_sample": true, "generation_max_new_tokens": 50, "generation_prompt_ratio": 0.5, "generation_temperature": 0.7, "gradient_accumulation_steps": 8, "gradient_checkpointing": true, "greater_is_better": false, "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "layer_offloading": false, "learning_rate": 5e-05, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": true, "load_in_4bit": false, "load_in_8bit": false, "local_rank": 0, "logging_steps": 10, "lora_dropout": 0.0, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "max_grad_norm": 1.0, "mean_resizing_embeddings": false, "merge_method": "memory_efficient", "metric_for_best_model": "eval_loss", "micro_batch_size": 2, "model_config_type": "qwen2", "num_epochs": 3.0, "num_generation_samples": 3, "optimizer": "adamw_bnb_8bit", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "kikiyaa/qwen-dpo-finetuned-ver2", "pretrain_multipack_attn": true, "profiler_steps_start": 0, "qlora_sharded_model_loading": false, "quantize_moe_experts": false, "ray_num_workers": 1, "resources_per_worker": { "GPU": 1 }, "rl": "dpo", "rl_beta": 0.1, "sample_packing": false, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 100, "save_total_limit": 3, "seed": 42, "sequence_len": 2048, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": true, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "Qwen/Qwen2.5-7B", "tokenizer_save_jinja_files": true, "tokenizer_type": "AutoTokenizer", "torch_dtype": "torch.bfloat16", "train_on_inputs": false, "trl": { "async_prefetch": false, "log_completions": false, "mask_truncated_completions": false, "ref_model_mixup_alpha": 0.9, "ref_model_sync_steps": 64, "replay_buffer_size": 0, "replay_recompute_logps": true, "reroll_max_groups": 1, "reroll_start_fraction": 1.0, "reward_num_workers": 1, "scale_rewards": true, "skip_zero_advantage_batches": true, "sync_ref_model": false, "use_data_producer": false, "use_vllm": false, "vllm_lora_sync": false, "vllm_server_host": "0.0.0.0", "vllm_server_port": 8000 }, "type_of_model": "AutoModelForCausalLM", "use_otel_metrics": false, "use_ray": false, "val_set_size": 0.05, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "warmup_ratio": 0.1, "weight_decay": 0.01, "world_size": 1 } [2026-04-16 10:31:57,826] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:307] [PID:135] EOS: 151643 / <|endoftext|> [2026-04-16 10:31:57,826] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:308] [PID:135] BOS: None / None [2026-04-16 10:31:57,826] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:309] [PID:135] PAD: 151643 / <|endoftext|> [2026-04-16 10:31:57,826] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:310] [PID:135] UNK: None / None [2026-04-16 10:31:57,831] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:475] [PID:135] Loading prepared dataset from disk at prepared_data/b845853b1ffa0f92dd415602f19f08af... [2026-04-16 10:31:57,902] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:135] loading tokenizer... Qwen/Qwen2.5-7B [2026-04-16 10:31:59,556] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:307] [PID:135] EOS: 151643 / <|endoftext|> [2026-04-16 10:31:59,556] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:308] [PID:135] BOS: None / None [2026-04-16 10:31:59,556] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:309] [PID:135] PAD: 151643 / <|endoftext|> [2026-04-16 10:31:59,556] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:310] [PID:135] UNK: None / None [2026-04-16 10:31:59,556] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:135] Loading model [2026-04-16 10:31:59,777] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:135] Patched OptimState8bit for torch.compile compatibility [2026-04-16 10:31:59,778] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:135] Patched OptimState4bit for torch.compile compatibility [2026-04-16 10:31:59,778] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:135] Patched OptimStateFp8 for torch.compile compatibility [2026-04-16 10:31:59,795] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:135] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-04-16 10:31:59,798] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:135] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-04-16 10:32:00,017] [INFO] [axolotl.loaders.patch_manager._apply_trl_trainer_utils_patches:852] [PID:135] Patched trl.trainer.utils with Triton selective_log_softmax and entropy_from_logits Fetching 4 files: 0%| | 0/4 [00:00