[2026-01-04 00:10:03,594] [DEBUG] [axolotl.utils.config.resolve_dtype:66] [PID:15692] bf16 support detected, enabling for this configuration. [2026-01-04 00:10:03,667] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:15692] baseline 0.000GB () [2026-01-04 00:10:03,668] [INFO] [axolotl.cli.config.load_cfg:259] [PID:15692] config: { "activation_offloading": true, "adapter": "lora", "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": 4, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_86", "fp8": false, "n_gpu": 1, "n_node": 1 }, "chat_template": "llama3", "context_parallel_size": 1, "cut_cross_entropy": true, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_num_proc": 96, "dataset_prepared_path": "dataset_prepareds", "datasets": [ { "chat_template": "tokenizer_default", "message_property_mappings": { "content": "content", "role": "role" }, "path": "WokeAI/polititune-tankie-warmup", "split": "train", "trust_remote_code": false, "type": "chat_template" } ], "ddp": false, "device": "cuda:0", "dion_rank_fraction": 1.0, "dion_rank_multiple_of": 1, "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": false, "eval_table_size": 0, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "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": 1e-05, "liger_fused_linear_cross_entropy": false, "liger_glu_activation": true, "liger_layer_norm": true, "liger_rms_norm": true, "liger_rope": true, "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": 16, "lora_dropout": 0.35, "lora_r": 64, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "rex", "max_grad_norm": 0.1, "mean_resizing_embeddings": false, "micro_batch_size": 2, "model_config_type": "llama", "num_epochs": 2.0, "optimizer": "adamw_torch_8bit", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "./output", "pad_to_sequence_len": true, "peft_use_rslora": 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, "ray_num_workers": 1, "remove_unused_columns": false, "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, "sequence_len": 4096, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "special_tokens": { "pad_token": "<|reserved_special_token_2|>" }, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": false, "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": { "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 }, "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": "newyear", "weight_decay": 0.0, "world_size": 1 } [2026-01-04 00:10:03,672] [WARNING] [axolotl.cli.checks.check_user_token:46] [PID:15692] Error verifying HuggingFace token. Remember to log in using `huggingface-cli login` and get your access token from https://huggingface.co/settings/tokens if you want to use gated models or datasets. [2026-01-04 00:10:04,197] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:15692] EOS: 128009 / <|eot_id|> [2026-01-04 00:10:04,198] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:15692] BOS: 128000 / <|begin_of_text|> [2026-01-04 00:10:04,198] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:15692] PAD: 128004 / <|reserved_special_token_2|> [2026-01-04 00:10:04,198] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:15692] UNK: None / None [2026-01-04 00:10:04,199] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:475] [PID:15692] Loading prepared dataset from disk at dataset_prepareds/a420619428aa6c5576289a496238883a... [2026-01-04 00:10:04,213] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:417] [PID:15692] total_num_tokens: 684_427 [2026-01-04 00:10:04,225] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:435] [PID:15692] `total_supervised_tokens: 498_319` [2026-01-04 00:10:04,242] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:15692] Using single process for pack_parallel, running sequentially. [2026-01-04 00:10:05,245] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:15692] Using single process for pack_parallel, running sequentially. [2026-01-04 00:10:05,463] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:15692] generate_batches time: 0.21780610084533691 [2026-01-04 00:10:05,463] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:15692] Using single process for pack_parallel, running sequentially. [2026-01-04 00:10:05,685] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:15692] generate_batches time: 0.22262859344482422 [2026-01-04 00:10:05,686] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:15692] Using single process for pack_parallel, running sequentially. [2026-01-04 00:10:05,930] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:15692] generate_batches time: 0.2447659969329834 [2026-01-04 00:10:05,931] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:15692] Using single process for pack_parallel, running sequentially. [2026-01-04 00:10:06,188] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:15692] generate_batches time: 0.25751280784606934 [2026-01-04 00:10:06,211] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:438] [PID:15692] gather_len_batches: [84] [2026-01-04 00:10:06,212] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:494] [PID:15692] data_loader_len: 42 [2026-01-04 00:10:06,212] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:510] [PID:15692] sample_packing_eff_est across ranks: [0.9946216401599702] [2026-01-04 00:10:06,212] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:522] [PID:15692] sample_packing_eff_est: 1.0 [2026-01-04 00:10:06,212] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:533] [PID:15692] total_num_steps: 84 [2026-01-04 00:10:06,212] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:15692] Maximum number of steps set at 84 [2026-01-04 00:10:06,240] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:15692] loading tokenizer... shb777/Llama-3.3-8B-Instruct-128K [2026-01-04 00:10:06,719] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:15692] EOS: 128009 / <|eot_id|> [2026-01-04 00:10:06,719] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:15692] BOS: 128000 / <|begin_of_text|> [2026-01-04 00:10:06,719] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:15692] PAD: 128004 / <|reserved_special_token_2|> [2026-01-04 00:10:06,719] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:15692] UNK: None / None [2026-01-04 00:10:06,719] [DEBUG] [axolotl.train.setup_model_and_tokenizer:82] [PID:15692] Loading model [2026-01-04 00:10:06,766] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:15692] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-01-04 00:10:06,767] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:15692] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-01-04 00:10:06,768] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:344] [PID:15692] Applying multipack dataloader patch for sample packing... [2026-01-04 00:10:06,873] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:98] [PID:15692] Applying LIGER to llama with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': False, 'rms_norm': True, 'swiglu': True} [2026-01-04 00:10:07,074] [INFO] [axolotl.integrations.cut_cross_entropy.pre_model_load:94] [PID:15692] Applying Cut Cross Entropy to model type: llama Loading checkpoint shards: 0%| | 0/4 [00:00