[2026-02-08 23:19:24,226] [WARNING] [axolotl.utils.trainer.prepare_optim_env:658] [PID:31858] P2P support not detected, setting `NCCL_P2P_DISABLE=1` [2026-02-08 23:19:24,227] [DEBUG] [axolotl.utils.config.resolve_dtype:66] [PID:31858] bf16 support detected, enabling for this configuration. [2026-02-08 23:19:24,357] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:31858] baseline 0.000GB () [2026-02-08 23:19:24,357] [INFO] [axolotl.cli.config.load_cfg:259] [PID:31858] config: { "activation_offloading": true, "axolotl_config_path": "train.yml", "base_model": "Nanbeige/Nanbeige4-3B-Thinking-2511", "base_model_config": "Nanbeige/Nanbeige4-3B-Thinking-2511", "batch_size": 16, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_89", "fp8": false, "n_gpu": 4, "n_node": 1 }, "chat_template": "tokenizer_default", "context_parallel_size": 1, "cosine_min_lr_ratio": 0.1, "dataloader_num_workers": 4, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_num_proc": 32, "dataset_prepared_path": "last_run_prepared", "datasets": [ { "chat_template": "tokenizer_default", "message_property_mappings": { "content": "content", "role": "role" }, "path": "WokeAI/polititune-tankie-warmup-3", "split": "train", "trust_remote_code": false, "type": "chat_template" } ], "ddp": true, "device": "cuda:0", "device_map": { "": 0 }, "dion_rank_fraction": 1.0, "dion_rank_multiple_of": 1, "eaft_alpha": 1.0, "eaft_k": 20, "env_capabilities": { "torch_version": "2.8.0" }, "eval_batch_size": 2, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": true, "eval_table_size": 0, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "fsdp": [ "auto_wrap", "full_shard" ], "fsdp_config": { "activation_checkpointing": true, "auto_wrap_policy": "TRANSFORMER_BASED_WRAP", "cpu_ram_efficient_loading": true, "fsdp_version": 2, "offload_params": false, "reshard_after_forward": true, "state_dict_type": "SHARDED_STATE_DICT", "transformer_layer_cls_to_wrap": "LlamaDecoderLayer" }, "fsdp_version": 2, "gradient_accumulation_steps": 2, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "group_by_length": false, "include_tkps": true, "is_llama_derived_model": true, "learning_rate": 5e-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_dropout": 0.0, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "constant", "max_grad_norm": 0.1, "mean_resizing_embeddings": false, "micro_batch_size": 2, "model_config_type": "llama", "num_epochs": 4.0, "optimizer": "adamw_torch_8bit", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "./model-output", "pad_to_sequence_len": true, "pretrain_multipack_attn": true, "profiler_steps_start": 0, "qlora_sharded_model_loading": false, "ray_num_workers": 1, "resources_per_worker": { "GPU": 1 }, "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.25, "saves_per_epoch": 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, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "Nanbeige/Nanbeige4-3B-Thinking-2511", "tokenizer_save_jinja_files": true, "torch_dtype": "torch.bfloat16", "train_on_inputs": false, "trl": { "log_completions": false, "mask_truncated_completions": false, "ref_model_mixup_alpha": 0.9, "ref_model_sync_steps": 64, "scale_rewards": true, "sync_ref_model": false, "use_vllm": false, "vllm_server_host": "0.0.0.0", "vllm_server_port": 8000 }, "trust_remote_code": false, "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": "polititune-3-wip-warmup", "warmup_ratio": 0.2, "weight_decay": 0.001, "world_size": 4 } [2026-02-08 23:19:25,836] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:31858] EOS: 166101 / <|im_end|> [2026-02-08 23:19:25,837] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:31858] BOS: 166100 / <|im_start|> [2026-02-08 23:19:25,838] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:31858] PAD: 0 / [2026-02-08 23:19:25,839] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:31858] UNK: 0 / [2026-02-08 23:19:25,842] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:475] [PID:31858] Loading prepared dataset from disk at last_run_prepared/22d86a2984af8d0d207033404eb271ea... [2026-02-08 23:19:25,851] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:417] [PID:31858] total_num_tokens: 999_041 [2026-02-08 23:19:25,861] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:435] [PID:31858] `total_supervised_tokens: 807_335` [2026-02-08 23:19:25,875] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:31858] Using single process for pack_parallel, running sequentially. [2026-02-08 23:19:26,396] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:31858] Using single process for pack_parallel, running sequentially. [2026-02-08 23:19:26,646] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:31858] generate_batches time: 0.2507505416870117 [2026-02-08 23:19:27,223] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:438] [PID:31858] gather_len_batches: [246, 246, 246, 246] [2026-02-08 23:19:27,225] [WARNING] [py.warnings._showwarnmsg:110] [PID:31858] /root/axolotl/.venv/lib/python3.12/site-packages/torch/distributed/distributed_c10d.py:4807: UserWarning: No device id is provided via `init_process_group` or `barrier `. Using the current device set by the user. warnings.warn( # warn only once [2026-02-08 23:19:27,291] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:494] [PID:31858] data_loader_len: 30 [2026-02-08 23:19:27,297] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:510] [PID:31858] sample_packing_eff_est across ranks: [0.9955366849899292, 0.991489827632904, 0.991489827632904, 0.991489827632904] [2026-02-08 23:19:27,297] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:522] [PID:31858] sample_packing_eff_est: 1.0 [2026-02-08 23:19:27,297] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:533] [PID:31858] total_num_steps: 120 [2026-02-08 23:19:27,297] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:31858] Maximum number of steps set at 120 [2026-02-08 23:19:27,340] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:31858] loading tokenizer... Nanbeige/Nanbeige4-3B-Thinking-2511 [2026-02-08 23:19:28,851] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:31858] EOS: 166101 / <|im_end|> [2026-02-08 23:19:28,852] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:31858] BOS: 166100 / <|im_start|> [2026-02-08 23:19:28,853] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:31858] PAD: 0 / [2026-02-08 23:19:28,854] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:31858] UNK: 0 / [2026-02-08 23:19:28,856] [DEBUG] [axolotl.train.setup_model_and_tokenizer:82] [PID:31858] Loading model [2026-02-08 23:19:28,933] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:31858] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-02-08 23:19:28,934] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:31858] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-02-08 23:19:28,935] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:338] [PID:31858] Applying multipack dataloader patch for sample packing... Loading weights: 0%| | 0/291 [00:00