[2025-10-04 06:10:32,094] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:1331] baseline 0.000GB () [2025-10-04 06:10:32,095] [INFO] [axolotl.cli.config.load_cfg:248] [PID:1331] config: { "activation_offloading": false, "adapter": "qlora", "axolotl_config_path": "bf16-co-sft.yaml", "base_model": "google/gemma-2-2b-it", "base_model_config": "google/gemma-2-2b-it", "batch_size": 4, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_90", "fp8": false, "n_gpu": 1, "n_node": 1 }, "context_parallel_size": 1, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_prepared_path": "last_run_prepared", "dataset_processes": 16, "datasets": [ { "chat_template": "jinja", "chat_template_jinja": "{{ bos_token }}\n{% set last = None %}\n{% for m in messages %}\n {% set raw_role = 'model' if m['role']=='assistant' else m['role'] %}\n {% set role = 'user' if raw_role=='system' else raw_role %}\n {% if role == last and role == 'user' %}\n {{ m['content'] | trim }}\n {% else %}\n {{ '' + role + '\\n' + m['content'] | trim + '\\n' }}\n {% endif %}\n {% set last = role %}\n{% endfor %}\n{% if add_generation_prompt %}\n{{ 'model\\n' }}\n{% endif %}\n", "data_files": "./co-sft-dataset.jsonl", "field_messages": "conversations", "message_property_mappings": { "content": "value", "role": "from" }, "path": ".", "roles_to_train": [ "assistant", "user" ], "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": "2.7.1" }, "eot_tokens": [ "" ], "eval_batch_size": 1, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": true, "eval_steps": 0.02, "eval_table_size": 0, "evals_per_epoch": 5, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "gradient_accumulation_steps": 4, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "hub_model_id": "AiAF/gemma-2-2b-it-co-sft-qlora", "hub_strategy": "checkpoint", "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "learning_rate": 0.0002, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": true, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 128, "lora_dropout": 0.05, "lora_r": 64, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "gemma2", "num_epochs": 10.0, "optimizer": "adamw_bnb_8bit", "output_dir": "./outputs/sft/gemma-2-2b-it-co", "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 }, "resume_from_checkpoint": "./outputs/sft/gemma-2-2b-it-co/checkpoint-15792", "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.02, "save_total_limit": 100, "saves_per_epoch": 5, "sequence_len": 2048, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "special_tokens": { "bos_token": "", "eos_token": "", "pad_token": "" }, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": true, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "google/gemma-2-2b-it", "tokenizer_save_jinja_files": true, "tokenizer_type": "AutoTokenizer", "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 }, "type_of_model": "AutoModelForCausalLM", "use_ray": false, "use_wandb": true, "val_set_size": 0.05, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_log_model": "false", "wandb_name": "gemma-2-2b-it_SFT-co_QLoRA", "wandb_project": "co-sft", "wandb_run_id": "1", "weight_decay": 0.0, "world_size": 1 } Aborted! [2025-10-04 06:16:33,481] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:1589] baseline 0.000GB () [2025-10-04 06:16:33,482] [INFO] [axolotl.cli.config.load_cfg:248] [PID:1589] config: { "activation_offloading": false, "adapter": "qlora", "axolotl_config_path": "bf16-co-sft.yaml", "base_model": "google/gemma-2-2b-it", "base_model_config": "google/gemma-2-2b-it", "batch_size": 4, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_90", "fp8": false, "n_gpu": 1, "n_node": 1 }, "context_parallel_size": 1, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_prepared_path": "last_run_prepared", "dataset_processes": 16, "datasets": [ { "chat_template": "jinja", "chat_template_jinja": "{{ bos_token }}\n{% set last = None %}\n{% for m in messages %}\n {% set raw_role = 'model' if m['role']=='assistant' else m['role'] %}\n {% set role = 'user' if raw_role=='system' else raw_role %}\n {% if role == last and role == 'user' %}\n {{ m['content'] | trim }}\n {% else %}\n {{ '' + role + '\\n' + m['content'] | trim + '\\n' }}\n {% endif %}\n {% set last = role %}\n{% endfor %}\n{% if add_generation_prompt %}\n{{ 'model\\n' }}\n{% endif %}\n", "data_files": "./co-sft-dataset.jsonl", "field_messages": "conversations", "message_property_mappings": { "content": "value", "role": "from" }, "path": ".", "roles_to_train": [ "assistant", "user" ], "trust_remote_code": false, "type": "chat_template" } ], "ddp": false, "device": "cuda:0", "device_map": "auto", "dion_rank_fraction": 1.0, "dion_rank_multiple_of": 1, "env_capabilities": { "torch_version": "2.7.1" }, "eot_tokens": [ "" ], "eval_batch_size": 1, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": true, "eval_steps": 0.02, "eval_table_size": 0, "evals_per_epoch": 5, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "gradient_accumulation_steps": 4, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "hub_model_id": "AiAF/gemma-2-2b-it-co-sft-qlora", "hub_strategy": "checkpoint", "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "is_preprocess": true, "learning_rate": 0.0002, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": true, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 128, "lora_dropout": 0.05, "lora_r": 64, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "gemma2", "num_epochs": 10.0, "optimizer": "adamw_bnb_8bit", "output_dir": "./outputs/sft/gemma-2-2b-it-co", "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 }, "resume_from_checkpoint": "./outputs/sft/gemma-2-2b-it-co/checkpoint-15792", "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.02, "save_total_limit": 100, "saves_per_epoch": 5, "sequence_len": 2048, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "special_tokens": { "bos_token": "", "eos_token": "", "pad_token": "" }, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": true, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "google/gemma-2-2b-it", "tokenizer_save_jinja_files": true, "tokenizer_type": "AutoTokenizer", "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 }, "type_of_model": "AutoModelForCausalLM", "use_ray": false, "use_wandb": true, "val_set_size": 0.05, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_log_model": "false", "wandb_name": "gemma-2-2b-it_SFT-co_QLoRA", "wandb_project": "co-sft", "wandb_run_id": "1", "weight_decay": 0.0, "world_size": 1 } tokenizer.json: 0%| | 0.00/17.5M [00:00 [2025-10-04 06:16:39,937] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:1589] BOS: 2 / [2025-10-04 06:16:39,937] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:1589] PAD: 0 / [2025-10-04 06:16:39,938] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:1589] UNK: 3 / [2025-10-04 06:16:39,939] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:476] [PID:1589] Unable to find prepared dataset in last_run_prepared/33f329184708d44d4d7e9c5bafe9d9a6 [2025-10-04 06:16:39,939] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:1589] Loading raw datasets... Generating train split: 0 examples [00:00, ? examples/s] Generating train split: 17944 examples [00:00, 150478.83 examples/s] Generating train split: 53494 examples [00:00, 183714.91 examples/s] Generating train split: 90651 examples [00:00, 200703.74 examples/s] Generating train split: 126559 examples [00:00, 203291.33 examples/s] Generating train split: 162816 examples [00:00, 209240.01 examples/s] Generating train split: 170296 examples [00:01, 89959.29 examples/s] [2025-10-04 06:16:50,205] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:1589] Loading dataset: . with base_type: chat_template and prompt_style: None [2025-10-04 06:16:50,229] [INFO] [axolotl.prompt_strategies.chat_template.__call__:969] [PID:1589] Using chat template: --- {{ bos_token }} {% set last = None %} {% for m in messages %} {% set raw_role = 'model' if m['role']=='assistant' else m['role'] %} {% set role = 'user' if raw_role=='system' else raw_role %} {% if role == last and role == 'user' %} {{ m['content'] | trim }} {% else %} {{ '' + role + '\n' + m['content'] | trim + '\n' }} {% endif %} {% set last = role %} {% endfor %} {% if add_generation_prompt %} {{ 'model\n' }} {% endif %} --- Tokenizing Prompts (num_proc=16): 0%| | 0/170296 [00:002048) (num_proc=16): 0%| | 0/170296 [00:002048) (num_proc=16): 1%|▍ | 1000/170296 [00:00<01:32, 1823.56 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 10%|██████▌ | 17000/170296 [00:00<00:04, 33892.99 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 20%|█████████████▏ | 34000/170296 [00:00<00:02, 62933.18 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 29%|███████████████████▍ | 50000/170296 [00:00<00:01, 83374.35 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 39%|█████████████████████████▌ | 67000/170296 [00:00<00:01, 102768.19 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 49%|████████████████████████████████ | 84000/170296 [00:01<00:00, 111671.90 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 59%|█████████████████████████████████████▌ | 100000/170296 [00:01<00:00, 121091.83 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 68%|███████████████████████████████████████████▏ | 115000/170296 [00:01<00:00, 127825.43 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 77%|█████████████████████████████████████████████████▏ | 131000/170296 [00:01<00:00, 135734.57 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 86%|███████████████████████████████████████████████████████▏ | 147000/170296 [00:01<00:00, 139867.27 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 96%|█████████████████████████████████████████████████████████████▍ | 163506/170296 [00:01<00:00, 145879.89 examples/s] Dropping Long Sequences (>2048) (num_proc=16): 100%|█████████████████████████████████████████████████████████████████| 170296/170296 [00:01<00:00, 89169.90 examples/s] [2025-10-04 06:17:47,510] [WARNING] [axolotl.utils.data.utils.handle_long_seq_in_dataset:260] [PID:1589] Dropped 8 samples from dataset Drop Samples with Zero Trainable Tokens (num_proc=16): 0%| | 0/170288 [00:00(-100, 2)  (-100, 108)  (-100, 141) (-100, 106) user(-100, 1645)  (-100, 108) You(-100, 2045)  are(-100, 708)  an(-100, 671)  anonymous(-100, 35466)  message(-100, 3969)  board(-100, 4924)  user(-100, 2425) .(-100, 235265)  You(-100, 1646)  are(-100, 708)  opinion(-100, 7837) ated(-100, 840)  and(-100, 578)  use(-100, 1281)  informal(-100, 35200)  language(-100, 5255) .(-100, 235265)  You(-100, 1646)  discuss(-100, 9742)  various(-100, 4282)  topics(-100, 16254) .(-100, 235265) (-100, 107)  (-100, 109)  (-100, 141) (-100, 106) user(-100, 1645)  (-100, 108) Season(27089, 27089)  (235248, 235248) 1(235274, 235274)  is(603, 603)  coming(5163, 5163) ,(235269, 235269)  fre(3956, 3956) ns(558, 558)  (108, 108) https(2468, 2468) ://(1336, 1336) www(1920, 1920) .(235265, 235265) youtube(11386, 11386) .(235265, 235265) com(872, 872) /(235283, 235283) watch(8170, 8170) ?(235336, 235336) v(235272, 235272) =(235293, 235293) Dn(94439, 94439) _(235298, 235298) rpt(132372, 132372) E(235291, 235291) _(235298, 235298) 3(235304, 235304) fk(46396, 46396) &(235343, 235343) amp(1657, 1657) ;(235289, 235289) ab(670, 670) _(235298, 235298) channel(11310, 11310) =(235293, 235293) Hannah(76993, 76993) Da(7735, 7735) igle(104474, 104474)  [(892, 892) Embed(62669, 62669) ](235307, 235307) (107, 107)  (-100, 109)  (-100, 141) (-100, 106) model(-100, 2516)  (-100, 108) Not(2440, 2440)  even(1693, 1693)  half(3933, 3933)  the(573, 573)  hype(54787, 54787)  and(578, 578)  the(573, 573)  stuff(6668, 6668)  posted(5727, 5727)  here(1517, 1517)  is(603, 603)  already(3303, 3303)  way(1703, 1703)  better(2525, 2525)  than(1178, 1178)  all(832, 832)  Hell(24598, 24598) uva(80337, 80337) Boss(43672, 43672) /(235283, 235283) Haz(26313, 26313) bin(3049, 3049) Hotel(23585, 23585)  fanart(110315, 110315)  and(578, 578)  porn(34163, 34163)  Lo(3107, 3107) ona(2977, 2977)  not(780, 780)  with(675, 675) standing(8898, 8898) (107, 107)  (-100, 109) [2025-10-04 06:17:53,941] [INFO] [axolotl.utils.tokenization.check_example_labels:45] [PID:1589] [2025-10-04 06:17:53,941] [INFO] [axolotl.utils.tokenization.check_example_labels:48] [PID:1589] Total input len: 114 [2025-10-04 06:17:53,941] [INFO] [axolotl.utils.tokenization.check_example_labels:49] [PID:1589] Count of labels: 74 [2025-10-04 06:17:53,941] [INFO] [axolotl.common.datasets.load_datasets:88] [PID:1589] printing prompters... [2025-10-04 06:17:53,941] [INFO] [axolotl.common.datasets.load_datasets:90] [PID:1589] Pre-tokenized or custom dataset types are unsupported for logging model.safetensors.index.json: 0%| | 0.00/24.2k [00:00' + role + '\\n' + m['content'] | trim + '\\n' }}\n {% endif %}\n {% set last = role %}\n{% endfor %}\n{% if add_generation_prompt %}\n{{ 'model\\n' }}\n{% endif %}\n", "data_files": "./co-sft-dataset.jsonl", "field_messages": "conversations", "message_property_mappings": { "content": "value", "role": "from" }, "path": ".", "roles_to_train": [ "assistant", "user" ], "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": "2.7.1" }, "eot_tokens": [ "" ], "eval_batch_size": 1, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": true, "eval_steps": 0.02, "eval_table_size": 0, "evals_per_epoch": 5, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "gradient_accumulation_steps": 4, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "hub_model_id": "AiAF/gemma-2-2b-it-co-sft-qlora", "hub_strategy": "checkpoint", "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "learning_rate": 0.0002, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": true, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 128, "lora_dropout": 0.05, "lora_r": 64, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "gemma2", "num_epochs": 10.0, "optimizer": "adamw_bnb_8bit", "output_dir": "./outputs/sft/gemma-2-2b-it-co", "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 }, "resume_from_checkpoint": "./outputs/sft/gemma-2-2b-it-co/checkpoint-15792", "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.02, "save_total_limit": 100, "saves_per_epoch": 5, "sequence_len": 2048, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "special_tokens": { "bos_token": "", "eos_token": "", "pad_token": "" }, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": true, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "google/gemma-2-2b-it", "tokenizer_save_jinja_files": true, "tokenizer_type": "AutoTokenizer", "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 }, "type_of_model": "AutoModelForCausalLM", "use_ray": false, "use_wandb": true, "val_set_size": 0.05, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_log_model": "false", "wandb_name": "gemma-2-2b-it_SFT-co_QLoRA", "wandb_project": "co-sft", "wandb_run_id": "1", "weight_decay": 0.0, "world_size": 1 } [2025-10-04 06:22:37,606] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:278] [PID:2522] EOS: 1 / [2025-10-04 06:22:37,606] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:2522] BOS: 2 / [2025-10-04 06:22:37,606] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:2522] PAD: 0 / [2025-10-04 06:22:37,607] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:2522] UNK: 3 / [2025-10-04 06:22:37,608] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:470] [PID:2522] Loading prepared dataset from disk at last_run_prepared/33f329184708d44d4d7e9c5bafe9d9a6... [2025-10-04 06:22:37,684] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:404] [PID:2522] total_num_tokens: 1_183_014 [2025-10-04 06:22:37,747] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:422] [PID:2522] `total_supervised_tokens: 842_414` [2025-10-04 06:22:39,790] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 0.9035689830780029 [2025-10-04 06:22:40,779] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 0.9887309074401855 [2025-10-04 06:22:41,808] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 1.0285992622375488 [2025-10-04 06:22:42,767] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 0.9584131240844727 [2025-10-04 06:22:42,808] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:434] [PID:2522] gather_len_batches: [583] [2025-10-04 06:22:42,808] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:481] [PID:2522] data_loader_len: 145 [2025-10-04 06:22:42,808] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:497] [PID:2522] sample_packing_eff_est across ranks: [0.9908122721912521] [2025-10-04 06:22:42,808] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:509] [PID:2522] sample_packing_eff_est: None [2025-10-04 06:22:42,808] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:520] [PID:2522] total_num_steps: 1450 [2025-10-04 06:22:42,987] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:404] [PID:2522] total_num_tokens: 22_892_853 [2025-10-04 06:22:44,177] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:422] [PID:2522] `total_supervised_tokens: 16_421_933` [2025-10-04 06:22:48,206] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 1.13698410987854 [2025-10-04 06:22:49,346] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 1.1380696296691895 [2025-10-04 06:22:50,485] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 1.136497974395752 [2025-10-04 06:22:51,381] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2522] generate_batches time: 0.8946337699890137 [2025-10-04 06:22:51,381] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:434] [PID:2522] gather_len_batches: [11268] [2025-10-04 06:22:51,381] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:481] [PID:2522] data_loader_len: 2817 [2025-10-04 06:22:51,381] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:497] [PID:2522] sample_packing_eff_est across ranks: [0.9920261695869942] [2025-10-04 06:22:51,382] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:509] [PID:2522] sample_packing_eff_est: 1.0 [2025-10-04 06:22:51,382] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:520] [PID:2522] total_num_steps: 28170 [2025-10-04 06:22:51,384] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:2522] Maximum number of steps set at 28170 [2025-10-04 06:22:51,446] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:2522] Loading tokenizer... google/gemma-2-2b-it [2025-10-04 06:22:52,570] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:278] [PID:2522] EOS: 1 / [2025-10-04 06:22:52,570] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:2522] BOS: 2 / [2025-10-04 06:22:52,570] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:2522] PAD: 0 / [2025-10-04 06:22:52,570] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:2522] UNK: 3 / [2025-10-04 06:22:52,570] [DEBUG] [axolotl.train.setup_model_and_tokenizer:79] [PID:2522] Loading model [2025-10-04 06:22:52,719] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:2522] Patched Trainer.evaluation_loop with nanmean loss calculation [2025-10-04 06:22:52,722] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:2522] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2025-10-04 06:22:52,723] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:301] [PID:2522] Applying multipack dataloader patch for sample packing... Loading checkpoint shards: 0%| | 0/2 [00:00", line 198, in _run_module_as_main File "", line 88, in _run_code File "/workspace/axolotl/src/axolotl/cli/train.py", line 121, in fire.Fire(do_cli) File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 135, in Fire component_trace = _Fire(component, args, parsed_flag_args, context, name) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 468, in _Fire component, remaining_args = _CallAndUpdateTrace( ^^^^^^^^^^^^^^^^^^^^ File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 684, in _CallAndUpdateTrace component = fn(*varargs, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/axolotl/src/axolotl/cli/train.py", line 88, in do_cli return do_train(parsed_cfg, parsed_cli_args) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/axolotl/src/axolotl/cli/train.py", line 45, in do_train model, tokenizer, trainer = train(cfg=cfg, dataset_meta=dataset_meta) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/axolotl/src/axolotl/train.py", line 583, in train execute_training(cfg, trainer, resume_from_checkpoint) File "/workspace/axolotl/src/axolotl/train.py", line 204, in execute_training trainer.train(resume_from_checkpoint=resume_from_checkpoint) File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/transformers/trainer.py", line 2319, in train return inner_training_loop( ^^^^^^^^^^^^^^^^^^^^ File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/transformers/trainer.py", line 2515, in _inner_training_loop self._load_scaler(resume_from_checkpoint) File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/transformers/trainer.py", line 3641, in _load_scaler self.accelerator.scaler.load_state_dict( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AttributeError: 'NoneType' object has no attribute 'load_state_dict' [2025-10-04 06:23:59,267] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:2952] baseline 0.000GB () [2025-10-04 06:23:59,268] [INFO] [axolotl.cli.config.load_cfg:248] [PID:2952] config: { "activation_offloading": false, "adapter": "qlora", "axolotl_config_path": "bf16-co-sft.yaml", "base_model": "google/gemma-2-2b-it", "base_model_config": "google/gemma-2-2b-it", "batch_size": 4, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_90", "fp8": false, "n_gpu": 1, "n_node": 1 }, "context_parallel_size": 1, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_prepared_path": "last_run_prepared", "dataset_processes": 16, "datasets": [ { "chat_template": "jinja", "chat_template_jinja": "{{ bos_token }}\n{% set last = None %}\n{% for m in messages %}\n {% set raw_role = 'model' if m['role']=='assistant' else m['role'] %}\n {% set role = 'user' if raw_role=='system' else raw_role %}\n {% if role == last and role == 'user' %}\n {{ m['content'] | trim }}\n {% else %}\n {{ '' + role + '\\n' + m['content'] | trim + '\\n' }}\n {% endif %}\n {% set last = role %}\n{% endfor %}\n{% if add_generation_prompt %}\n{{ 'model\\n' }}\n{% endif %}\n", "data_files": "./co-sft-dataset.jsonl", "field_messages": "conversations", "message_property_mappings": { "content": "value", "role": "from" }, "path": ".", "roles_to_train": [ "assistant", "user" ], "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": "2.7.1" }, "eot_tokens": [ "" ], "eval_batch_size": 1, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": true, "eval_steps": 0.02, "eval_table_size": 0, "evals_per_epoch": 5, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "gradient_accumulation_steps": 4, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "hub_model_id": "AiAF/gemma-2-2b-it-co-sft-qlora", "hub_strategy": "checkpoint", "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "learning_rate": 0.0002, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": true, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 128, "lora_dropout": 0.05, "lora_r": 64, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "gemma2", "num_epochs": 10.0, "optimizer": "adamw_bnb_8bit", "output_dir": "./outputs/sft/gemma-2-2b-it-co", "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 }, "resume_from_checkpoint": "./outputs/sft/gemma-2-2b-it-co/checkpoint-15792", "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.02, "save_total_limit": 100, "saves_per_epoch": 5, "sequence_len": 2048, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "special_tokens": { "bos_token": "", "eos_token": "", "pad_token": "" }, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": true, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "google/gemma-2-2b-it", "tokenizer_save_jinja_files": true, "tokenizer_type": "AutoTokenizer", "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 }, "type_of_model": "AutoModelForCausalLM", "use_ray": false, "use_wandb": true, "val_set_size": 0.05, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_log_model": "false", "wandb_name": "gemma-2-2b-it_SFT-co_QLoRA", "wandb_project": "co-sft", "wandb_run_id": "1", "weight_decay": 0.0, "world_size": 1 } [2025-10-04 06:24:01,209] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:278] [PID:2952] EOS: 1 / [2025-10-04 06:24:01,209] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:2952] BOS: 2 / [2025-10-04 06:24:01,209] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:2952] PAD: 0 / [2025-10-04 06:24:01,209] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:2952] UNK: 3 / [2025-10-04 06:24:01,211] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:470] [PID:2952] Loading prepared dataset from disk at last_run_prepared/33f329184708d44d4d7e9c5bafe9d9a6... [2025-10-04 06:24:01,287] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:404] [PID:2952] total_num_tokens: 1_183_014 [2025-10-04 06:24:01,351] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:422] [PID:2952] `total_supervised_tokens: 842_414` [2025-10-04 06:24:03,441] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 0.9338798522949219 [2025-10-04 06:24:04,449] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.0082221031188965 [2025-10-04 06:24:05,459] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.009068489074707 [2025-10-04 06:24:06,494] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.034280776977539 [2025-10-04 06:24:06,535] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:434] [PID:2952] gather_len_batches: [583] [2025-10-04 06:24:06,535] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:481] [PID:2952] data_loader_len: 145 [2025-10-04 06:24:06,535] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:497] [PID:2952] sample_packing_eff_est across ranks: [0.9908122721912521] [2025-10-04 06:24:06,535] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:509] [PID:2952] sample_packing_eff_est: None [2025-10-04 06:24:06,535] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:520] [PID:2952] total_num_steps: 1450 [2025-10-04 06:24:06,715] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:404] [PID:2952] total_num_tokens: 22_892_853 [2025-10-04 06:24:07,902] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:422] [PID:2952] `total_supervised_tokens: 16_421_933` [2025-10-04 06:24:12,762] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.3197410106658936 [2025-10-04 06:24:14,045] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.2806613445281982 [2025-10-04 06:24:15,354] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.3064305782318115 [2025-10-04 06:24:16,675] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:2952] generate_batches time: 1.3178660869598389 [2025-10-04 06:24:16,675] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:434] [PID:2952] gather_len_batches: [11267] [2025-10-04 06:24:16,675] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:481] [PID:2952] data_loader_len: 2816 [2025-10-04 06:24:16,675] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:497] [PID:2952] sample_packing_eff_est across ranks: [0.9919381381583325] [2025-10-04 06:24:16,676] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:509] [PID:2952] sample_packing_eff_est: 1.0 [2025-10-04 06:24:16,676] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:520] [PID:2952] total_num_steps: 28160 [2025-10-04 06:24:16,681] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:2952] Maximum number of steps set at 28160 [2025-10-04 06:24:16,785] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:2952] Loading tokenizer... google/gemma-2-2b-it [2025-10-04 06:24:18,595] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:278] [PID:2952] EOS: 1 / [2025-10-04 06:24:18,595] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:2952] BOS: 2 / [2025-10-04 06:24:18,595] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:2952] PAD: 0 / [2025-10-04 06:24:18,595] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:2952] UNK: 3 / [2025-10-04 06:24:18,595] [DEBUG] [axolotl.train.setup_model_and_tokenizer:79] [PID:2952] Loading model [2025-10-04 06:24:18,805] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:2952] Patched Trainer.evaluation_loop with nanmean loss calculation [2025-10-04 06:24:18,808] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:2952] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2025-10-04 06:24:18,809] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:301] [PID:2952] Applying multipack dataloader patch for sample packing... Loading checkpoint shards: 0%| | 0/2 [00:00", line 198, in _run_module_as_main File "", line 88, in _run_code File "/workspace/axolotl/src/axolotl/cli/train.py", line 121, in fire.Fire(do_cli) File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 135, in Fire component_trace = _Fire(component, args, parsed_flag_args, context, name) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 468, in _Fire component, remaining_args = _CallAndUpdateTrace( ^^^^^^^^^^^^^^^^^^^^ File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 684, in _CallAndUpdateTrace component = fn(*varargs, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/axolotl/src/axolotl/cli/train.py", line 88, in do_cli return do_train(parsed_cfg, parsed_cli_args) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/axolotl/src/axolotl/cli/train.py", line 45, in do_train model, tokenizer, trainer = train(cfg=cfg, dataset_meta=dataset_meta) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/axolotl/src/axolotl/train.py", line 583, in train execute_training(cfg, trainer, resume_from_checkpoint) File "/workspace/axolotl/src/axolotl/train.py", line 204, in execute_training trainer.train(resume_from_checkpoint=resume_from_checkpoint) File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/transformers/trainer.py", line 2319, in train return inner_training_loop( ^^^^^^^^^^^^^^^^^^^^ File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/transformers/trainer.py", line 2515, in _inner_training_loop self._load_scaler(resume_from_checkpoint) File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/transformers/trainer.py", line 3641, in _load_scaler self.accelerator.scaler.load_state_dict( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AttributeError: 'NoneType' object has no attribute 'load_state_dict' [2025-10-04 06:33:49,280] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:4425] baseline 0.000GB () [2025-10-04 06:33:49,282] [INFO] [axolotl.cli.config.load_cfg:248] [PID:4425] config: { "activation_offloading": false, "adapter": "qlora", "axolotl_config_path": "bf16-co-sft.yaml", "base_model": "google/gemma-2-2b-it", "base_model_config": "google/gemma-2-2b-it", "batch_size": 4, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_90", "fp8": false, "n_gpu": 1, "n_node": 1 }, "context_parallel_size": 1, "dataloader_num_workers": 1, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_prepared_path": "last_run_prepared", "dataset_processes": 16, "datasets": [ { "chat_template": "jinja", "chat_template_jinja": "{{ bos_token }}\n{% set last = None %}\n{% for m in messages %}\n {% set raw_role = 'model' if m['role']=='assistant' else m['role'] %}\n {% set role = 'user' if raw_role=='system' else raw_role %}\n {% if role == last and role == 'user' %}\n {{ m['content'] | trim }}\n {% else %}\n {{ '' + role + '\\n' + m['content'] | trim + '\\n' }}\n {% endif %}\n {% set last = role %}\n{% endfor %}\n{% if add_generation_prompt %}\n{{ 'model\\n' }}\n{% endif %}\n", "data_files": "./co-sft-dataset.jsonl", "field_messages": "conversations", "message_property_mappings": { "content": "value", "role": "from" }, "path": ".", "roles_to_train": [ "assistant", "user" ], "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": "2.7.1" }, "eot_tokens": [ "" ], "eval_batch_size": 1, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_sample_packing": true, "eval_steps": 0.02, "eval_table_size": 0, "evals_per_epoch": 5, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "gradient_accumulation_steps": 4, "gradient_checkpointing": true, "gradient_checkpointing_kwargs": { "use_reentrant": true }, "hub_model_id": "AiAF/gemma-2-2b-it-co-sft-qlora", "hub_strategy": "checkpoint", "include_tkps": true, "is_falcon_derived_model": false, "is_llama_derived_model": false, "is_mistral_derived_model": false, "learning_rate": 0.0002, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": true, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 128, "lora_dropout": 0.05, "lora_r": 64, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "gemma2", "num_epochs": 10.0, "optimizer": "adamw_bnb_8bit", "output_dir": "./outputs/sft/gemma-2-2b-it-co", "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 }, "resume_from_checkpoint": "outputs/sft/gemma-2-2b-it-co/checkpoint-15792/", "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.02, "save_total_limit": 100, "saves_per_epoch": 5, "sequence_len": 2048, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "special_tokens": { "bos_token": "", "eos_token": "", "pad_token": "" }, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tf32": true, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "google/gemma-2-2b-it", "tokenizer_save_jinja_files": true, "tokenizer_type": "AutoTokenizer", "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 }, "type_of_model": "AutoModelForCausalLM", "use_ray": false, "use_wandb": true, "val_set_size": 0.05, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_log_model": "false", "wandb_name": "gemma-2-2b-it_SFT-co_QLoRA", "wandb_project": "co-sft", "wandb_run_id": "1", "weight_decay": 0.0, "world_size": 1 } [2025-10-04 06:33:50,937] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:278] [PID:4425] EOS: 1 / [2025-10-04 06:33:50,937] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:4425] BOS: 2 / [2025-10-04 06:33:50,937] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:4425] PAD: 0 / [2025-10-04 06:33:50,937] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:4425] UNK: 3 / [2025-10-04 06:33:50,939] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:470] [PID:4425] Loading prepared dataset from disk at last_run_prepared/33f329184708d44d4d7e9c5bafe9d9a6... [2025-10-04 06:33:51,015] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:404] [PID:4425] total_num_tokens: 1_183_014 [2025-10-04 06:33:51,077] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:422] [PID:4425] `total_supervised_tokens: 842_414` [2025-10-04 06:33:53,136] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 0.9459962844848633 [2025-10-04 06:33:54,143] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 1.0062956809997559 [2025-10-04 06:33:55,127] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 0.9836509227752686 [2025-10-04 06:33:56,060] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 0.9326572418212891 [2025-10-04 06:33:56,101] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:434] [PID:4425] gather_len_batches: [583] [2025-10-04 06:33:56,102] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:481] [PID:4425] data_loader_len: 145 [2025-10-04 06:33:56,102] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:497] [PID:4425] sample_packing_eff_est across ranks: [0.9908122721912521] [2025-10-04 06:33:56,102] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:509] [PID:4425] sample_packing_eff_est: None [2025-10-04 06:33:56,102] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:520] [PID:4425] total_num_steps: 1450 [2025-10-04 06:33:56,281] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:404] [PID:4425] total_num_tokens: 22_892_853 [2025-10-04 06:33:57,496] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:422] [PID:4425] `total_supervised_tokens: 16_421_933` [2025-10-04 06:34:02,073] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 1.3816823959350586 [2025-10-04 06:34:03,388] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 1.3118422031402588 [2025-10-04 06:34:04,792] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 1.4012658596038818 [2025-10-04 06:34:06,147] [DEBUG] [axolotl.utils.samplers.multipack.__len__:458] [PID:4425] generate_batches time: 1.3523709774017334 [2025-10-04 06:34:06,148] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:434] [PID:4425] gather_len_batches: [11268] [2025-10-04 06:34:06,148] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:481] [PID:4425] data_loader_len: 2817 [2025-10-04 06:34:06,148] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:497] [PID:4425] sample_packing_eff_est across ranks: [0.9921142166420742] [2025-10-04 06:34:06,148] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:509] [PID:4425] sample_packing_eff_est: 1.0 [2025-10-04 06:34:06,148] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:520] [PID:4425] total_num_steps: 28170 [2025-10-04 06:34:06,153] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:4425] Maximum number of steps set at 28170 [2025-10-04 06:34:06,247] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:4425] Loading tokenizer... google/gemma-2-2b-it [2025-10-04 06:34:07,925] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:278] [PID:4425] EOS: 1 / [2025-10-04 06:34:07,925] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:279] [PID:4425] BOS: 2 / [2025-10-04 06:34:07,925] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:4425] PAD: 0 / [2025-10-04 06:34:07,925] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:4425] UNK: 3 / [2025-10-04 06:34:07,926] [DEBUG] [axolotl.train.setup_model_and_tokenizer:79] [PID:4425] Loading model [2025-10-04 06:34:08,084] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:4425] Patched Trainer.evaluation_loop with nanmean loss calculation [2025-10-04 06:34:08,087] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:4425] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2025-10-04 06:34:08,088] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:301] [PID:4425] Applying multipack dataloader patch for sample packing... Loading checkpoint shards: 0%| | 0/2 [00:00