basic_50m_times_4 / wandb /debug.log
eoinf's picture
Upload folder using huggingface_hub
e52422e verified
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_setup.py:_flush():80] Current SDK version is 0.21.1
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_setup.py:_flush():80] Configure stats pid to 387
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_setup.py:_flush():80] Loading settings from /root/.config/wandb/settings
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_setup.py:_flush():80] Loading settings from /notebooks/toy_models/models/gelu_2l_v7_50m_subset/wandb/settings
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_setup.py:_flush():80] Loading settings from environment variables
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_init.py:setup_run_log_directory():703] Logging user logs to /notebooks/toy_models/models/gelu_2l_v7_50m_subset/wandb/run-20250819_121927-k06awv3c/logs/debug.log
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_init.py:setup_run_log_directory():704] Logging internal logs to /notebooks/toy_models/models/gelu_2l_v7_50m_subset/wandb/run-20250819_121927-k06awv3c/logs/debug-internal.log
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_init.py:init():830] calling init triggers
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_init.py:init():835] wandb.init called with sweep_config: {}
config: {'model_name': 'gelu_2l_50m_subset', 'n_layers': 2, 'd_model': 512, 'd_mlp': 2048, 'd_head': 64, 'n_heads': 8, 'attn_only': False, 'layer_norm_eps': 1e-05, 'init_range': 0.02, 'n_ctx': 1024, 'd_vocab': 48262, 'dataset_name': 'eoinf/c4-code-test-50m', 'tokenizer_name': 'NeelNanda/gpt-neox-tokenizer-digits', 'seed': 10, 'device': 'cuda', 'use_bfloat16_matmul': False, 'batch_size_per_device': 32, 'n_devices': 1, 'batches_per_step': 1, 'max_tokens': 200000000, 'lr_hidden': 0.002, 'lr_vector': 0.001, 'lr_schedule': 'constant_with_warmup', 'warmup_tokens': 30000000, 'weight_decay': 0.05, 'grad_norm_clip': 1.0, 'train_loss_moving_average_beta': 0.99, 'log_interval': 25, 'save_checkpoints': True, 'checkpoint_interval': 500, 'checkpoint_interval_ratio': 1.08, 'save_log_checkpoints': True, 'use_wandb': True, 'batch_size': 32, 'tokens_per_step': 32768, 'warmup_steps': 915, 'max_steps': 6103, '_wandb': {}}
2025-08-19 12:19:27,434 INFO MainThread:387 [wandb_init.py:init():871] starting backend
2025-08-19 12:19:27,906 INFO MainThread:387 [wandb_init.py:init():874] sending inform_init request
2025-08-19 12:19:27,933 INFO MainThread:387 [wandb_init.py:init():882] backend started and connected
2025-08-19 12:19:27,934 INFO MainThread:387 [wandb_init.py:init():953] updated telemetry
2025-08-19 12:19:27,940 INFO MainThread:387 [wandb_init.py:init():977] communicating run to backend with 90.0 second timeout
2025-08-19 12:19:28,473 INFO MainThread:387 [wandb_init.py:init():1029] starting run threads in backend
2025-08-19 12:19:29,237 INFO MainThread:387 [wandb_run.py:_console_start():2494] atexit reg
2025-08-19 12:19:29,237 INFO MainThread:387 [wandb_run.py:_redirect():2342] redirect: wrap_raw
2025-08-19 12:19:29,238 INFO MainThread:387 [wandb_run.py:_redirect():2411] Wrapping output streams.
2025-08-19 12:19:29,238 INFO MainThread:387 [wandb_run.py:_redirect():2434] Redirects installed.
2025-08-19 12:19:29,248 INFO MainThread:387 [wandb_init.py:init():1075] run started, returning control to user process
2025-08-19 13:28:02,197 INFO MainThread:387 [wandb_run.py:_finish():2260] finishing run eoin/toy-transformer-replication/k06awv3c
2025-08-19 13:28:02,202 INFO MainThread:387 [wandb_run.py:_atexit_cleanup():2459] got exitcode: 0
2025-08-19 13:28:02,202 INFO MainThread:387 [wandb_run.py:_restore():2441] restore
2025-08-19 13:28:02,202 INFO MainThread:387 [wandb_run.py:_restore():2447] restore done
2025-08-19 13:28:02,673 INFO MainThread:387 [wandb_run.py:_footer_history_summary_info():3895] rendering history
2025-08-19 13:28:02,674 INFO MainThread:387 [wandb_run.py:_footer_history_summary_info():3927] rendering summary
2025-08-19 13:28:02,674 INFO MainThread:387 [wandb_run.py:_footer_sync_info():3856] logging synced files