[2026-05-16 18:20:43,308] [DEBUG] [axolotl.utils.config.resolve_dtype:74] [PID:14603] bf16 support detected, enabling for this configuration. [2026-05-16 18:20:43,452] [WARNING] [huggingface_hub.utils._http._warn_on_warning_headers:904] [PID:14603] Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads. [2026-05-16 18:20:43,465] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:14603] baseline 0.000GB () [2026-05-16 18:20:43,467] [INFO] [axolotl.cli.config.load_cfg:333] [PID:14603] config: { "activation_offloading": true, "attn_implementation": "flash_attention_2", "attn_needs_dtype_cast": true, "attn_supports_packing": true, "attn_uses_flash_lib": true, "axolotl_config_path": "train.yml", "base_model": "shb777/Llama-3.3-8B-Instruct-128K", "base_model_config": "shb777/Llama-3.3-8B-Instruct-128K", "batch_size": 32, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_80", "fp8": false, "n_gpu": 1, "n_node": 1, "tf32": true }, "chat_template": "tokenizer_default", "context_parallel_size": 1, "cosine_min_lr_ratio": 0.1, "cut_cross_entropy": true, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_num_proc": 32, "dataset_prepared_path": "last_run_prepared", "datasets": [ { "ds_type": "parquet", "message_property_mappings": { "content": "content", "role": "role" }, "path": "tankieV2-llama3.3.parquet", "trust_remote_code": false } ], "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": "2.8.0" }, "eval_batch_size": 4, "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, "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, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "group_by_length": false, "include_tkps": true, "is_llama_derived_model": true, "layer_offloading": false, "learning_rate": 5e-06, "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_with_warmup", "max_grad_norm": 0.1, "mean_resizing_embeddings": false, "merge_method": "memory_efficient", "micro_batch_size": 4, "model_config_type": "llama", "num_epochs": 3.0, "num_generation_samples": 3, "optimizer": "adamw_torch_fused", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "./model-output", "pad_to_sequence_len": true, "plugins": [ "axolotl.integrations.liger.LigerPlugin", "axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin" ], "pretrain_multipack_attn": true, "profiler_steps_start": 0, "qlora_sharded_model_loading": false, "quantize_moe_experts": false, "ray_num_workers": 1, "relora_prune_method": "magnitude", "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, "saves_per_epoch": 0, "sequence_len": 4096, "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": "shb777/Llama-3.3-8B-Instruct-128K", "tokenizer_save_jinja_files": true, "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 }, "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": "tankieV2-llama3.3", "warmup_ratio": 0.05, "weight_decay": 0.001, "world_size": 1 } [2026-05-16 18:20:43,470] [WARNING] [axolotl.cli.checks.check_user_token:48] [PID:14603] Error verifying HuggingFace token. Remember to log in using `hf auth login` and get your access token from https://huggingface.co/settings/tokens if you want to use gated models or datasets. [2026-05-16 18:20:45,130] [WARNING] [axolotl.loaders.tokenizer.load_tokenizer:305] [PID:14603] Tokenizer does not have a pad_token, falling back to eos_token: <|eot_id|> [2026-05-16 18:20:45,131] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:14603] EOS: 128009 / <|eot_id|> [2026-05-16 18:20:45,131] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:14603] BOS: 128000 / <|begin_of_text|> [2026-05-16 18:20:45,131] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:14603] PAD: 128009 / <|eot_id|> [2026-05-16 18:20:45,132] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:14603] UNK: None / None [2026-05-16 18:20:45,133] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:477] [PID:14603] Loading prepared dataset from disk at last_run_prepared/8e4debe9138a048d05e5d68d1585e372... [2026-05-16 18:20:45,154] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:420] [PID:14603] total_num_tokens: 1_063_789 [2026-05-16 18:20:45,170] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:438] [PID:14603] `total_supervised_tokens: 884_389` [2026-05-16 18:20:47,287] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:14603] generate_batches time: 1.073195457458496 [2026-05-16 18:20:48,338] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:14603] generate_batches time: 1.0501418113708496 [2026-05-16 18:20:49,358] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:14603] generate_batches time: 1.0195682048797607 [2026-05-16 18:20:50,431] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:14603] generate_batches time: 1.0725367069244385 [2026-05-16 18:20:50,460] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:438] [PID:14603] gather_len_batches: [66] [2026-05-16 18:20:50,460] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:495] [PID:14603] data_loader_len: 8 [2026-05-16 18:20:50,460] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:504] [PID:14603] sample_packing_eff_est across ranks: [0.9837655732125947] [2026-05-16 18:20:50,460] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:516] [PID:14603] sample_packing_eff_est: 0.99 [2026-05-16 18:20:50,460] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:521] [PID:14603] total_num_steps: 24 [2026-05-16 18:20:50,461] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:14603] Maximum number of steps set at 24 [2026-05-16 18:20:50,514] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:14603] loading tokenizer... shb777/Llama-3.3-8B-Instruct-128K [2026-05-16 18:20:52,167] [WARNING] [axolotl.loaders.tokenizer.load_tokenizer:305] [PID:14603] Tokenizer does not have a pad_token, falling back to eos_token: <|eot_id|> [2026-05-16 18:20:52,167] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:14603] EOS: 128009 / <|eot_id|> [2026-05-16 18:20:52,168] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:14603] BOS: 128000 / <|begin_of_text|> [2026-05-16 18:20:52,168] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:14603] PAD: 128009 / <|eot_id|> [2026-05-16 18:20:52,168] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:14603] UNK: None / None [2026-05-16 18:20:52,168] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:14603] Loading model [2026-05-16 18:20:52,254] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:14603] Patched OptimState8bit for torch.compile compatibility [2026-05-16 18:20:52,254] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:14603] Patched OptimState4bit for torch.compile compatibility [2026-05-16 18:20:52,255] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:14603] Patched OptimStateFp8 for torch.compile compatibility [2026-05-16 18:20:52,262] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:14603] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-05-16 18:20:52,264] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:14603] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-05-16 18:20:52,265] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:598] [PID:14603] Applying multipack dataloader patch for sample packing... [2026-05-16 18:20:52,504] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:104] [PID:14603] Applying LIGER to llama with kwargs: {'rope': None, 'cross_entropy': None, 'fused_linear_cross_entropy': None, 'rms_norm': None, 'swiglu': None} [2026-05-16 18:20:52,534] [INFO] [axolotl.integrations.cut_cross_entropy.pre_model_load:94] [PID:14603] Applying Cut Cross Entropy to model type: llama Loading weights: 0%| | 0/291 [00:00