[2026-03-04 15:11:19,031] [DEBUG] [axolotl.utils.config.resolve_dtype:66] [PID:2633803] bf16 support detected, enabling for this configuration. [2026-03-04 15:11:19,338] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:2633803] baseline 0.000GB () [2026-03-04 15:11:19,338] [INFO] [axolotl.cli.config.load_cfg:259] [PID:2633803] config: { "activation_offloading": false, "axolotl_config_path": "ood_test.yaml", "base_model": "kakaocorp/kanana-1.5-2.1b-instruct-2505", "base_model_config": "kakaocorp/kanana-1.5-2.1b-instruct-2505", "batch_size": 48, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_89", "fp8": false, "n_gpu": 3, "n_node": 1 }, "context_parallel_size": 1, "dataloader_num_workers": 32, "dataset_num_proc": 64, "dataset_prepared_path": "preprocess", "datasets": [ { "chat_template": "tokenizer_default", "message_property_mappings": { "content": "content", "role": "role" }, "path": "train_final_aug100.jsonl", "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": 16, "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, "fsdp": [ "full_shard", "auto_wrap" ], "fsdp_config": { "activation_checkpointing": true, "state_dict_type": "FULL_STATE_DICT", "transformer_layer_cls_to_wrap": "LlamaDecoderLayer" }, "gradient_accumulation_steps": 1, "gradient_checkpointing": false, "hub_model_id": "CaffeineThief/ttp_sft_kanana-1.5_100per_ood_added_anno", "include_tkps": true, "is_llama_derived_model": true, "learning_rate": 2e-05, "liger_fused_linear_cross_entropy": true, "liger_glu_activation": 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_dropout": 0.0, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "mean_resizing_embeddings": false, "micro_batch_size": 16, "model_config_type": "llama", "num_epochs": 2.0, "optimizer": "adamw_torch_fused", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "./outputs-ood100_added_anno", "pad_to_sequence_len": false, "plugins": [ "axolotl.integrations.liger.LigerPlugin" ], "pretrain_multipack_attn": true, "profiler_steps_start": 0, "qlora_sharded_model_loading": false, "ray_num_workers": 1, "resources_per_worker": { "GPU": 1 }, "sample_packing": false, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.5, "saves_per_epoch": 1, "sequence_len": 6000, "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": "kakaocorp/kanana-1.5-2.1b-instruct-2505", "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_name": "CaffeineThief/ttp_sft_kanana-1.5_100per_ood_added_anno", "wandb_project": "TTP_SFT_LLM_RE", "warmup_ratio": 0.05, "weight_decay": 0.01, "world_size": 3 } [2026-03-04 15:11:21,159] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:2633803] EOS: 128009 / <|eot_id|> [2026-03-04 15:11:21,159] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:2633803] BOS: 128000 / <|begin_of_text|> [2026-03-04 15:11:21,159] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:2633803] PAD: 128001 / <|end_of_text|> [2026-03-04 15:11:21,159] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:2633803] UNK: None / None [2026-03-04 15:11:43,066] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:475] [PID:2633803] Loading prepared dataset from disk at preprocess/f749c854d17e07cecfdf92462a84250c... Loading dataset from disk: 0%| | 0/64 [00:00 [2026-03-04 15:11:45,193] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:2633803] BOS: 128000 / <|begin_of_text|> [2026-03-04 15:11:45,193] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:2633803] PAD: 128001 / <|end_of_text|> [2026-03-04 15:11:45,193] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:2633803] UNK: None / None [2026-03-04 15:11:45,193] [DEBUG] [axolotl.train.setup_model_and_tokenizer:82] [PID:2633803] Loading model [2026-03-04 15:11:45,447] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:2633803] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-03-04 15:11:45,451] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:2633803] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-03-04 15:11:45,574] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:98] [PID:2633803] Applying LIGER to llama with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': True, 'rms_norm': True, 'swiglu': True} [2026-03-04 15:11:46,150] [INFO] [axolotl.loaders.model._configure_embedding_dtypes:351] [PID:2633803] Converting modules to torch.bfloat16 [2026-03-04 15:11:46,156] [DEBUG] [axolotl.loaders.model.log_gpu_memory_usage:127] [PID:2633803] Memory usage after model load 0.000GB () [2026-03-04 15:11:52,872] [INFO] [axolotl.train.save_initial_configs:406] [PID:2633803] Pre-saving tokenizer to ./outputs-ood100_added_anno... [2026-03-04 15:11:53,069] [INFO] [axolotl.train.save_initial_configs:411] [PID:2633803] Pre-saving model config to ./outputs-ood100_added_anno... [2026-03-04 15:11:53,072] [INFO] [axolotl.train.execute_training:207] [PID:2633803] Starting trainer... [2026-03-04 15:11:56,706] [WARNING] [py.warnings._showwarnmsg:109] [PID:2633803] /usr/local/lib/python3.10/dist-packages/accelerate/accelerator.py:1992: UserWarning: Upcasted low precision parameters in LlamaForCausalLM because mixed precision turned on in FSDP. Affects: model.embed_tokens.weight, model.norm.weight. warnings.warn( [2026-03-04 15:11:56,706] [WARNING] [py.warnings._showwarnmsg:109] [PID:2633803] /usr/local/lib/python3.10/dist-packages/accelerate/accelerator.py:1992: UserWarning: Upcasted low precision parameters in LlamaDecoderLayer because mixed precision turned on in FSDP. Affects: self_attn.q_proj.weight, self_attn.k_proj.weight, self_attn.v_proj.weight, self_attn.o_proj.weight, mlp.gate_proj.weight, mlp.up_proj.weight, mlp.down_proj.weight, input_layernorm.weight, post_attention_layernorm.weight. warnings.warn( [2026-03-04 15:11:56,706] [WARNING] [py.warnings._showwarnmsg:109] [PID:2633803] /usr/local/lib/python3.10/dist-packages/accelerate/accelerator.py:1998: UserWarning: FSDP upcast of low precision parameters may affect the precision of model checkpoints. warnings.warn( wandb: [wandb.login()] Loaded credentials for https://api.wandb.ai from /root/.netrc. wandb: Currently logged in as: menbu3344 (menbu3344-kumoh-national-institute-of-technology) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin wandb: ⢿ setting up run kx1e7d7b (0.0s) m wandb: ⣻ setting up run kx1e7d7b (0.0s) m wandb: ⣽ setting up run kx1e7d7b (0.0s) m wandb: ⣾ setting up run kx1e7d7b (0.0s) m wandb: ⣷ setting up run kx1e7d7b (0.0s) m wandb: Tracking run with wandb version 0.25.0 wandb: Run data is saved locally in /data4/khs/organize/ood_test/wandb/run-20260304_151157-kx1e7d7b wandb: Run `wandb offline` to turn off syncing. wandb: Syncing run CaffeineThief/ttp_sft_kanana-1.5_100per_ood_added_anno wandb: ⭐️ View project at https://wandb.ai/menbu3344-kumoh-national-institute-of-technology/TTP_SFT_LLM_RE wandb: 🚀 View run at https://wandb.ai/menbu3344-kumoh-national-institute-of-technology/TTP_SFT_LLM_RE/runs/kx1e7d7b wandb: Detected [huggingface_hub.inference] in use. wandb: Use W&B Weave for improved LLM call tracing. Install Weave with `pip install weave` then add `import weave` to the top of your script. wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/ wandb: WARNING Saving files without folders. If you want to preserve subdirectories pass base_path to wandb.save, i.e. wandb.save("/mnt/folder/file.h5", base_path="/mnt") wandb: WARNING Symlinked 1 file into the W&B run directory; call wandb.save again to sync new files. [2026-03-04 15:12:01,022] [INFO] [axolotl.utils.callbacks.on_train_begin:757] [PID:2633803] The Axolotl config has been saved to the WandB run under files. 0%| | 0/1025 [00:00