Hugging Face's logo

yuccaaa
/
nas

TensorBoard
Safetensors
Model card Files
xet
Metrics Community
nas
File size: 7,934 Bytes
052f594
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
 
2025-06-18 20:58:00,417 INFO    MainThread:7254 [wandb_setup.py:_flush():81] Current SDK version is 0.20.1
2025-06-18 20:58:00,417 INFO    MainThread:7254 [wandb_setup.py:_flush():81] Configure stats pid to 7254
2025-06-18 20:58:00,417 INFO    MainThread:7254 [wandb_setup.py:_flush():81] Loading settings from /root/.config/wandb/settings
2025-06-18 20:58:00,417 INFO    MainThread:7254 [wandb_setup.py:_flush():81] Loading settings from /nas/shared/kilab/wangyujia/EasyR1/examples/wandb/settings
2025-06-18 20:58:00,417 INFO    MainThread:7254 [wandb_setup.py:_flush():81] Loading settings from environment variables
2025-06-18 20:58:00,417 INFO    MainThread:7254 [wandb_init.py:setup_run_log_directory():703] Logging user logs to /nas/shared/kilab/wangyujia/EasyR1/examples/wandb/run-20250618_205800-gokgllcc/logs/debug.log
2025-06-18 20:58:00,418 INFO    MainThread:7254 [wandb_init.py:setup_run_log_directory():704] Logging internal logs to /nas/shared/kilab/wangyujia/EasyR1/examples/wandb/run-20250618_205800-gokgllcc/logs/debug-internal.log
2025-06-18 20:58:00,418 INFO    MainThread:7254 [wandb_init.py:init():831] calling init triggers
2025-06-18 20:58:00,418 INFO    MainThread:7254 [wandb_init.py:init():836] wandb.init called with sweep_config: {}
config: {'data': {'train_files': '/nas/shared/kilab/wangyujia/rl_data/deeplocmulti@train', 'val_files': '/nas/shared/kilab/wangyujia/rl_data/deeplocmulti@validation', 'prompt_key': 'question', 'answer_key': 'answer', 'image_key': 'images', 'image_dir': None, 'max_prompt_length': 4096, 'max_response_length': 16384, 'rollout_batch_size': 128, 'val_batch_size': 256, 'format_prompt': '/nas/shared/kilab/wangyujia/EasyR1/examples/format_prompt/bio_format.jinja', 'override_chat_template': None, 'shuffle': True, 'seed': 1, 'min_pixels': 262144, 'max_pixels': 4194304, 'filter_overlong_prompts': True}, 'worker': {'hybrid_engine': True, 'actor': {'strategy': 'fsdp', 'global_batch_size': 64, 'micro_batch_size_per_device_for_update': 2, 'micro_batch_size_per_device_for_experience': 16, 'max_grad_norm': 1.0, 'clip_ratio_low': 0.2, 'clip_ratio_high': 0.3, 'clip_ratio_dual': 3.0, 'ppo_epochs': 1, 'padding_free': True, 'ulysses_sequence_parallel_size': 1, 'use_torch_compile': True, 'model': {'model_path': '/oss/wangyujia/BIO/pretrain_output/qwen2.5-7b-instruct-bio/bio_all/save1epoch/checkpoint-1300', 'tokenizer_path': '/oss/wangyujia/BIO/pretrain_output/qwen2.5-7b-instruct-bio/bio_all/save1epoch/checkpoint-1300', 'override_config': {}, 'enable_gradient_checkpointing': True, 'trust_remote_code': False, 'freeze_vision_tower': False}, 'optim': {'lr': 1e-06, 'betas': [0.9, 0.999], 'weight_decay': 0.01, 'strategy': 'adamw', 'lr_warmup_ratio': 0.0, 'min_lr_ratio': None, 'warmup_style': 'constant', 'training_steps': 31}, 'fsdp': {'enable_full_shard': True, 'enable_cpu_offload': False, 'enable_rank0_init': True, 'use_orig_params': False, 'torch_dtype': None, 'fsdp_size': -1, 'mp_param_dtype': 'bf16', 'mp_reduce_dtype': 'fp32', 'mp_buffer_dtype': 'fp32'}, 'offload': {'offload_params': True, 'offload_optimizer': True}, 'global_batch_size_per_device': -1, 'disable_kl': False, 'use_kl_loss': True, 'kl_penalty': 'low_var_kl', 'kl_coef': 0.01}, 'critic': {'strategy': 'fsdp', 'global_batch_size': 256, 'micro_batch_size_per_device_for_update': 4, 'micro_batch_size_per_device_for_experience': 16, 'max_grad_norm': 1.0, 'cliprange_value': 0.5, 'ppo_epochs': 1, 'padding_free': False, 'ulysses_sequence_parallel_size': 1, 'model': {'model_path': None, 'tokenizer_path': None, 'override_config': {}, 'enable_gradient_checkpointing': True, 'trust_remote_code': True, 'freeze_vision_tower': False}, 'optim': {'lr': 1e-06, 'betas': [0.9, 0.999], 'weight_decay': 0.01, 'strategy': 'adamw', 'lr_warmup_ratio': 0.0, 'min_lr_ratio': None, 'warmup_style': 'constant', 'training_steps': 31}, 'fsdp': {'enable_full_shard': True, 'enable_cpu_offload': False, 'enable_rank0_init': True, 'use_orig_params': False, 'torch_dtype': None, 'fsdp_size': -1, 'mp_param_dtype': 'bf16', 'mp_reduce_dtype': 'fp32', 'mp_buffer_dtype': 'fp32'}, 'offload': {'offload_params': False, 'offload_optimizer': False}, 'global_batch_size_per_device': -1}, 'ref': {'strategy': 'fsdp', 'fsdp': {'enable_full_shard': True, 'enable_cpu_offload': True, 'enable_rank0_init': True, 'use_orig_params': False, 'torch_dtype': None, 'fsdp_size': -1, 'mp_param_dtype': 'bf16', 'mp_reduce_dtype': 'fp32', 'mp_buffer_dtype': 'fp32'}, 'offload': {'offload_params': False, 'offload_optimizer': False}, 'micro_batch_size_per_device_for_experience': 16, 'padding_free': True, 'ulysses_sequence_parallel_size': 1, 'use_torch_compile': True}, 'reward': {'reward_type': 'batch', 'reward_function': '/nas/shared/kilab/wangyujia/EasyR1/examples/reward_function/bio.py', 'reward_function_kwargs': {}, 'skip_special_tokens': True, 'num_cpus': 1, 'reward_function_name': 'compute_score'}, 'rollout': {'name': 'vllm', 'n': 5, 'temperature': 1.0, 'top_p': 0.99, 'top_k': -1, 'seed': 1, 'limit_images': 0, 'dtype': 'bf16', 'gpu_memory_utilization': 0.6, 'ignore_eos': False, 'enforce_eager': False, 'enable_chunked_prefill': False, 'tensor_parallel_size': 1, 'max_model_len': None, 'max_num_batched_tokens': 24576, 'disable_log_stats': True, 'val_override_config': {'temperature': 0.5, 'n': 1}, 'prompt_length': 4096, 'response_length': 16384, 'trust_remote_code': False}}, 'algorithm': {'gamma': 1.0, 'lam': 1.0, 'adv_estimator': 'grpo', 'disable_kl': False, 'use_kl_loss': True, 'kl_penalty': 'low_var_kl', 'kl_coef': 0.01, 'kl_type': 'fixed', 'kl_horizon': 0.0, 'kl_target': 0.0}, 'trainer': {'total_epochs': 1, 'max_steps': None, 'project_name': 'easy_r1', 'experiment_name': 'qwen2.5_7b_bio_06182042', 'logger': ['console', 'wandb'], 'nnodes': 1, 'n_gpus_per_node': 8, 'critic_warmup': 0, 'val_freq': 5, 'val_before_train': True, 'val_only': False, 'val_generations_to_log': 3, 'save_freq': 5, 'save_limit': 3, 'save_checkpoint_path': '/oss/wangyujia/BIO/rl/qwen2.5_7b_bio_06182042', 'load_checkpoint_path': None}, '_wandb': {}}
2025-06-18 20:58:00,418 INFO    MainThread:7254 [wandb_init.py:init():872] starting backend
2025-06-18 20:58:00,630 INFO    MainThread:7254 [wandb_init.py:init():875] sending inform_init request
2025-06-18 20:58:00,633 INFO    MainThread:7254 [wandb_init.py:init():883] backend started and connected
2025-06-18 20:58:00,637 INFO    MainThread:7254 [wandb_init.py:init():956] updated telemetry
2025-06-18 20:58:00,638 INFO    MainThread:7254 [wandb_init.py:init():980] communicating run to backend with 90.0 second timeout
2025-06-18 20:58:02,224 INFO    MainThread:7254 [wandb_init.py:init():1032] starting run threads in backend
2025-06-18 20:58:02,474 INFO    MainThread:7254 [wandb_run.py:_console_start():2453] atexit reg
2025-06-18 20:58:02,474 INFO    MainThread:7254 [wandb_run.py:_redirect():2301] redirect: wrap_raw
2025-06-18 20:58:02,474 INFO    MainThread:7254 [wandb_run.py:_redirect():2370] Wrapping output streams.
2025-06-18 20:58:02,474 INFO    MainThread:7254 [wandb_run.py:_redirect():2393] Redirects installed.
2025-06-18 20:58:02,481 INFO    MainThread:7254 [wandb_init.py:init():1078] run started, returning control to user process
2025-06-19 01:29:20,221 INFO    MainThread:7254 [wandb_run.py:_finish():2219] finishing run gia0603yucca/easy_r1/gokgllcc
2025-06-19 01:29:20,240 INFO    MainThread:7254 [wandb_run.py:_atexit_cleanup():2418] got exitcode: 0
2025-06-19 01:29:20,246 INFO    MainThread:7254 [wandb_run.py:_restore():2400] restore
2025-06-19 01:29:20,247 INFO    MainThread:7254 [wandb_run.py:_restore():2406] restore done
2025-06-19 01:29:22,887 INFO    MainThread:7254 [wandb_run.py:_footer_history_summary_info():4000] rendering history
2025-06-19 01:29:22,889 INFO    MainThread:7254 [wandb_run.py:_footer_history_summary_info():4032] rendering summary
2025-06-19 01:29:22,890 INFO    MainThread:7254 [wandb_run.py:_footer_sync_info():3961] logging synced files