[2026-04-21 05:20:16,043] [DEBUG] [axolotl.utils.config.resolve_dtype:74] [PID:12538] bf16 support detected, enabling for this configuration. [2026-04-21 05:20:16,046] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:12538] baseline 0.000GB () [2026-04-21 05:20:16,048] [INFO] [axolotl.cli.config.load_cfg:341] [PID:12538] config: { "activation_offloading": false, "adapter": "lora", "axolotl_config_path": "finetuning/axolotl/configs/qwen_32b/qwen_32b_em_unpop_mmlu/hf_qwen_32b_em_unpop_mmlu_1.yaml", "base_model": "models/hf_qwen_32b_em_unpop_1/merged", "base_model_config": "models/hf_qwen_32b_em_unpop_1/merged", "batch_size": 8, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_90", "fp8": true, "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": 12, "datasets": [ { "chat_template": "tokenizer_default", "message_property_mappings": { "content": "content", "role": "role" }, "path": "data/finetuning/mmlu/mmlu_sft_merged.jsonl", "roles": { "assistant": [ "assistant" ], "system": [ "system" ], "user": [ "user" ] }, "trust_remote_code": false, "type": "chat_template" } ], "ddp": false, "device": "cuda:0", "dion_rank_fraction": 1.0, "dion_rank_multiple_of": 1, "do_bench_eval": false, "eaft_alpha": 1.0, "eaft_k": 20, "env_capabilities": { "torch_version": "2.10.0" }, "eval_batch_size": 4, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": false, "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": 2, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": false }, "group_by_length": false, "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "layer_offloading": false, "learning_rate": 1e-05, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": false, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 64, "lora_dropout": 0.0, "lora_embedding_kernel": true, "lora_fan_in_fan_out": false, "lora_mlp_kernel": true, "lora_o_kernel": true, "lora_qkv_kernel": true, "lora_r": 32, "lora_target_modules": [ "q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj" ], "loraplus_lr_embedding": 1e-06, "lr_scheduler": "linear", "mean_resizing_embeddings": false, "merge_method": "memory_efficient", "micro_batch_size": 4, "model_config_type": "qwen2", "num_epochs": 1.0, "num_generation_samples": 3, "optimizer": "adamw_8bit", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "models/hf_qwen_32b_em_unpop_mmlu_1", "pad_to_sequence_len": false, "peft_use_dora": false, "peft_use_rslora": true, "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_beta": 0.1, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "saves_per_epoch": 1, "seed": 1, "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": false, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "models/hf_qwen_32b_em_unpop_1/merged", "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, "use_wandb": true, "val_set_size": 0.0, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_project": "hf_qwen_32b_em_unpop_mmlu_1", "warmup_steps": 10, "weight_decay": 0.01, "world_size": 1 } [2026-04-21 05:20:17,079] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:307] [PID:12538] EOS: 151645 / <|im_end|> [2026-04-21 05:20:17,079] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:308] [PID:12538] BOS: None / None [2026-04-21 05:20:17,079] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:309] [PID:12538] PAD: 151665 / <|PAD_TOKEN|> [2026-04-21 05:20:17,079] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:310] [PID:12538] UNK: None / None [2026-04-21 05:20:17,079] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:480] [PID:12538] Unable to find prepared dataset in last_run_prepared/4a7375e6a6d7cb0bffe2cae00ed9b9aa [2026-04-21 05:20:17,080] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:12538] Loading raw datasets... [2026-04-21 05:20:17,080] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:12538] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. [2026-04-21 05:20:17,497] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:12538] Loading dataset: data/finetuning/mmlu/mmlu_sft_merged.jsonl with base_type: chat_template and prompt_style: None [2026-04-21 05:20:17,499] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1168] [PID:12538] Using chat template: --- {%- if tools %} {{- '<|im_start|>system\n' }} {%- if messages[0]['role'] == 'system' %} {{- messages[0]['content'] }} {%- else %} {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }} {%- endif %} {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }} {%- for tool in tools %} {{- "\n" }} {{- tool | tojson }} {%- endfor %} {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }} {%- else %} {%- if messages[0]['role'] == 'system' %} {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }} {%- else %} {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }} {%- endif %} {%- endif %} {%- for message in messages %} {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %} {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }} {%- elif message.role == "assistant" %} {{- '<|im_start|>' + message.role }} {%- if message.content %} {{- '\n' + message.content }} {%- endif %} {%- for tool_call in message.tool_calls %} {%- if tool_call.function is defined %} {%- set tool_call = tool_call.function %} {%- endif %} {{- '\n\n{"name": "' }} {{- tool_call.name }} {{- '", "arguments": ' }} {{- tool_call.arguments | tojson }} {{- '}\n' }} {%- endfor %} {{- '<|im_end|>\n' }} {%- elif message.role == "tool" %} {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %} {{- '<|im_start|>user' }} {%- endif %} {{- '\n\n' }} {{- message.content }} {{- '\n' }} {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %} {{- '<|im_end|>\n' }} {%- endif %} {%- endif %} {%- endfor %} {%- if add_generation_prompt %} {{- '<|im_start|>assistant\n' }} {%- endif %} --- Tokenizing Prompts (num_proc=12): 0%| | 0/1912 [00:002048) (num_proc=12): 0%| | 0/1912 [00:002048) (num_proc=12): 8%|█▉ | 160/1912 [00:00<00:02, 706.26 examples/s] Dropping Invalid Sequences (2048) (num_proc=12): 100%|█████████████████████| 1912/1912 [00:00<00:00, 4907.96 examples/s] Saving the dataset (0/7 shards): 0%| | 0/1912 [00:00 [2026-04-21 05:20:44,720] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:308] [PID:12538] BOS: None / None [2026-04-21 05:20:44,720] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:309] [PID:12538] PAD: 151665 / <|PAD_TOKEN|> [2026-04-21 05:20:44,720] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:310] [PID:12538] UNK: None / None [2026-04-21 05:20:44,720] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:12538] Loading model [2026-04-21 05:20:44,722] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:12538] Patched OptimState8bit for torch.compile compatibility [2026-04-21 05:20:44,722] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:12538] Patched OptimState4bit for torch.compile compatibility [2026-04-21 05:20:44,722] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:12538] Patched OptimStateFp8 for torch.compile compatibility [2026-04-21 05:20:44,727] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:12538] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-04-21 05:20:44,728] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:12538] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-04-21 05:20:44,797] [INFO] [axolotl.monkeypatch.attention.flash_attn_4.patch_flash_attn_4:52] [PID:12538] Flash Attention 4 is available for your GPU and offers faster training speeds. To enable: pip install flash-attn-4 [2026-04-21 05:20:46,209] [INFO] [axolotl.monkeypatch.lora_kernels.patch_self_attn_lora:364] [PID:12538] Patched attention class with LoRA optims: Qwen2Attention Loading weights: 0%| | 0/771 [00:00