======================== START TIME: Tue Jul 2 16:05:32 UTC 2024 python3 version = Python 3.10.14 ======================== The token has not been saved to the git credentials helper. Pass `add_to_git_credential=True` in this function directly or `--add-to-git-credential` if using via `huggingface-cli` if you want to set the git credential as well. Token is valid (permission: write). Your token has been saved to /admin/home/ferdinand_mom/.cache/huggingface/token Login successful Already on 'bench_cluster' M examples/config_tiny_llama.py M examples/config_tiny_llama.yaml M examples/train_tiny_llama.sh M src/nanotron/models/llama.py M src/nanotron/trainer.py Your branch is up to date with 'origin/bench_cluster'. Job status: RUNNING W0702 16:05:37.682000 139819419109184 torch/distributed/run.py:757] W0702 16:05:37.682000 139819419109184 torch/distributed/run.py:757] ***************************************** W0702 16:05:37.682000 139819419109184 torch/distributed/run.py:757] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. W0702 16:05:37.682000 139819419109184 torch/distributed/run.py:757] ***************************************** W0702 16:05:37.718000 140455586768704 torch/distributed/run.py:757] W0702 16:05:37.718000 140455586768704 torch/distributed/run.py:757] ***************************************** W0702 16:05:37.718000 140455586768704 torch/distributed/run.py:757] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. W0702 16:05:37.718000 140455586768704 torch/distributed/run.py:757] ***************************************** [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Config: [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Config(general=GeneralArgs(project='bench_cluster', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: run='%date_%jobid', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: seed=42, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: step=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: consumed_train_samples=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: benchmark_csv_path=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: ignore_sanity_checks=True), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: parallelism=ParallelismArgs(dp=2, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: pp=8, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tp=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: pp_engine=, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tp_mode=, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tp_linear_async_communication=False, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: expert_parallel_size=1), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: model=ModelArgs(model_config=LlamaConfig(bos_token_id=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: eos_token_id=2, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: hidden_act='silu', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: hidden_size=2048, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: initializer_range=0.02, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: intermediate_size=4096, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: is_llama_config=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: max_position_embeddings=4096, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_attention_heads=32, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_hidden_layers=24, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_key_value_heads=32, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: pad_token_id=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: pretraining_tp=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: rms_norm_eps=1e-05, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: rope_scaling=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: rope_theta=10000.0, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tie_word_embeddings=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: use_cache=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: vocab_size=50257), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: init_method=RandomInit(std=0.025), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: dtype=torch.bfloat16, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: make_vocab_size_divisible_by=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: ddp_bucket_cap_mb=25), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tokenizer=TokenizerArgs(tokenizer_name_or_path='openai-community/gpt2', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tokenizer_revision=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tokenizer_max_length=None), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: checkpoints=CheckpointsArgs(checkpoints_path=Path('/dev/null'), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: checkpoint_interval=100000, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: save_initial_state=False, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: resume_checkpoint_path=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: checkpoints_path_is_shared_file_system=False), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: logging=LoggingArgs(log_level='info', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: log_level_replica='info', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: iteration_step_info_interval=1), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tokens=TokensArgs(sequence_length=4096, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: train_steps=20, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: micro_batch_size=4, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: batch_accumulation_per_replica=128, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: val_check_interval=-1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: limit_val_batches=0, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: limit_test_batches=0), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: optimizer=OptimizerArgs(optimizer_factory=AdamWOptimizerArgs(adam_eps=1e-08, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: adam_beta1=0.9, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: adam_beta2=0.95, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: torch_adam_is_fused=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: name='adamW'), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: zero_stage=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: weight_decay=0.01, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: clip_grad=1.0, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: accumulate_grad_in_fp32=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: learning_rate_scheduler=LRSchedulerArgs(learning_rate=0.0001, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: lr_warmup_steps=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: lr_warmup_style='linear', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: lr_decay_style='linear', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: lr_decay_steps=19, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: lr_decay_starting_step=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: min_decay_lr=1e-05)), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: data_stages=[DatasetStageArgs(name='Training Stage', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: start_training_step=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: data=DataArgs(dataset=PretrainDatasetsArgs(hf_dataset_or_datasets='roneneldan/TinyStories', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: hf_dataset_splits='train', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: hf_dataset_config_name=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: dataset_processing_num_proc_per_process=64, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: dataset_overwrite_cache=False, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: text_column_name='text'), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: seed=42, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_loading_workers=32))], [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: profiler=ProfilerArgs(profiler_export_path=Path('/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-1_pp-8_mbz-4')), [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: lighteval=None) [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Model Config: [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: LlamaConfig(bos_token_id=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: eos_token_id=2, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: hidden_act='silu', [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: hidden_size=2048, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: initializer_range=0.02, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: intermediate_size=4096, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: is_llama_config=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: max_position_embeddings=4096, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_attention_heads=32, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_hidden_layers=24, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: num_key_value_heads=32, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: pad_token_id=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: pretraining_tp=1, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: rms_norm_eps=1e-05, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: rope_scaling=None, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: rope_theta=10000.0, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: tie_word_embeddings=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: use_cache=True, [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: vocab_size=50257) [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Building model.. [default0]:07/02/2024 16:06:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Setting PP block ranks... [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Total number of parameters: 1.21G (2312.82MiB) [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Local number of parameters: 271M (516.35MiB) [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [After model building] Memory usage: 520.36MiB. Peak allocated: 522.39MiB Peak reserved: 534.00MiB [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Parametrizing model parameters using StandardParametrizator [default6]:07/02/2024 16:06:14 [INFO|DP=0|PP=3|TP=0|ip-26-0-163-43]: Local number of parameters: 168M (320.03MiB) [default6]:07/02/2024 16:06:14 [INFO|DP=0|PP=3|TP=0|ip-26-0-163-43]: [After model building] Memory usage: 324.04MiB. Peak allocated: 326.07MiB Peak reserved: 336.00MiB [default6]:07/02/2024 16:06:14 [INFO|DP=0|PP=3|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default4]:07/02/2024 16:06:14 [INFO|DP=0|PP=2|TP=0|ip-26-0-163-43]: Local number of parameters: 126M (240.02MiB) [default4]:07/02/2024 16:06:14 [INFO|DP=0|PP=2|TP=0|ip-26-0-163-43]: [After model building] Memory usage: 243.03MiB. Peak allocated: 245.06MiB Peak reserved: 262.00MiB [default4]:07/02/2024 16:06:14 [INFO|DP=0|PP=2|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default2]:07/02/2024 16:06:14 [INFO|DP=0|PP=1|TP=0|ip-26-0-163-43]: Local number of parameters: 126M (240.02MiB) [default2]:07/02/2024 16:06:14 [INFO|DP=0|PP=1|TP=0|ip-26-0-163-43]: [After model building] Memory usage: 243.03MiB. Peak allocated: 245.06MiB Peak reserved: 262.00MiB [default2]:07/02/2024 16:06:14 [INFO|DP=0|PP=1|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default4]:07/02/2024 16:06:14 [INFO|DP=0|PP=6|TP=0|ip-26-0-169-207]: Local number of parameters: 168M (320.03MiB) [default4]:07/02/2024 16:06:14 [INFO|DP=0|PP=6|TP=0|ip-26-0-169-207]: [After model building] Memory usage: 324.04MiB. Peak allocated: 326.07MiB Peak reserved: 336.00MiB [default4]:07/02/2024 16:06:14 [INFO|DP=0|PP=6|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=4|TP=0|ip-26-0-169-207]: Local number of parameters: 126M (240.02MiB) [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=4|TP=0|ip-26-0-169-207]: [After model building] Memory usage: 243.03MiB. Peak allocated: 245.06MiB Peak reserved: 262.00MiB [default0]:07/02/2024 16:06:14 [INFO|DP=0|PP=4|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default6]:07/02/2024 16:06:14 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: Local number of parameters: 103M (196.32MiB) [default6]:07/02/2024 16:06:14 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: [After model building] Memory usage: 196.33MiB. Peak allocated: 196.34MiB Peak reserved: 200.00MiB [default6]:07/02/2024 16:06:14 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default2]:07/02/2024 16:06:14 [INFO|DP=0|PP=5|TP=0|ip-26-0-169-207]: Local number of parameters: 126M (240.02MiB) [default2]:07/02/2024 16:06:14 [INFO|DP=0|PP=5|TP=0|ip-26-0-169-207]: [After model building] Memory usage: 243.03MiB. Peak allocated: 245.06MiB Peak reserved: 262.00MiB [default2]:07/02/2024 16:06:14 [INFO|DP=0|PP=5|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default5]:07/02/2024 16:06:14 [INFO|DP=1|PP=2|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default3]:07/02/2024 16:06:14 [INFO|DP=1|PP=1|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default7]:07/02/2024 16:06:14 [INFO|DP=1|PP=3|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default1]:07/02/2024 16:06:14 [INFO|DP=1|PP=0|TP=0|ip-26-0-163-43]: No checkpoint path provided. [default5]:07/02/2024 16:06:14 [INFO|DP=1|PP=6|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default7]:07/02/2024 16:06:14 [INFO|DP=1|PP=7|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default1]:07/02/2024 16:06:14 [INFO|DP=1|PP=4|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default3]:07/02/2024 16:06:14 [INFO|DP=1|PP=5|TP=0|ip-26-0-169-207]: No checkpoint path provided. [default0]:07/02/2024 16:06:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [Optimizer Building] Using LearningRateForSP as learning rate [default0]:07/02/2024 16:06:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [ZeRO sharding] Size of optimizer params per rank: [default0]:07/02/2024 16:06:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [ZeRO sharding] DP Rank 0 has 135M out of 271M (50.00%) params' optimizer states [default0]:07/02/2024 16:06:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [ZeRO sharding] DP Rank 1 has 135M out of 271M (50.00%) params' optimizer states [default0]:07/02/2024 16:06:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [Training Plan] Stage Training Stage has 19 remaining training steps and has consumed 0 samples [default0]:07/02/2024 16:06:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Using `datasets` library [default0]:07/02/2024 16:06:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Loading tokenizer from openai-community/gpt2 and transformers/hf_hub versions ('4.41.2', '0.23.4') [default0]:07/02/2024 16:06:18 [WARNING|DP=0|PP=0|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default0]:Repo card metadata block was not found. Setting CardData to empty. [default0]:07/02/2024 16:06:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [Training Plan] There are 1 training stages [default0]:07/02/2024 16:06:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [Stage Training Stage] start from step 1 [default0]:07/02/2024 16:06:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [default0]:07/02/2024 16:06:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: [Start training] datetime: 2024-07-02 16:06:20.739681 | mbs: 4 | grad_accum: 128 | global_batch_size: 1024 | sequence_length: 4096 | train_steps: 20 | start_iteration_step: 0 | consumed_train_samples: 0 [default0]:07/02/2024 16:06:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Resuming training from stage Training Stage, it has trained for 0 samples and has 19 remaining train steps [default0]:07/02/2024 16:06:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 2069.40MiB. Peak allocated 2069.40MiB. Peak reserved: 2086.00MiB [default5]:07/02/2024 16:06:20 [WARNING|DP=1|PP=2|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default5]:Repo card metadata block was not found. Setting CardData to empty. [default6]:07/02/2024 16:06:20 [WARNING|DP=0|PP=3|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default1]:Repo card metadata block was not found. Setting CardData to empty. [default4]:07/02/2024 16:06:20 [WARNING|DP=0|PP=2|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default3]:07/02/2024 16:06:20 [WARNING|DP=1|PP=1|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default1]:07/02/2024 16:06:20 [WARNING|DP=1|PP=0|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default2]:07/02/2024 16:06:20 [WARNING|DP=0|PP=1|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default7]:07/02/2024 16:06:20 [WARNING|DP=1|PP=3|TP=0|ip-26-0-163-43]: Repo card metadata block was not found. Setting CardData to empty. [default2]:Repo card metadata block was not found. Setting CardData to empty. [default6]:Repo card metadata block was not found. Setting CardData to empty. [default3]:Repo card metadata block was not found. Setting CardData to empty. [default0]:07/02/2024 16:06:20 [WARNING|DP=0|PP=4|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default5]:07/02/2024 16:06:20 [WARNING|DP=1|PP=6|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default1]:Repo card metadata block was not found. Setting CardData to empty. [default7]:Repo card metadata block was not found. Setting CardData to empty. [default4]:07/02/2024 16:06:20 [WARNING|DP=0|PP=6|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default7]:Repo card metadata block was not found. Setting CardData to empty. [default7]:07/02/2024 16:06:20 [WARNING|DP=1|PP=7|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default4]:Repo card metadata block was not found. Setting CardData to empty. [default1]:07/02/2024 16:06:20 [WARNING|DP=1|PP=4|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default3]:07/02/2024 16:06:20 [WARNING|DP=1|PP=5|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default4]:Repo card metadata block was not found. Setting CardData to empty. [default0]:Repo card metadata block was not found. Setting CardData to empty. [default3]:Repo card metadata block was not found. Setting CardData to empty. [default5]:Repo card metadata block was not found. Setting CardData to empty. [default6]:07/02/2024 16:06:20 [WARNING|DP=0|PP=7|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default2]:07/02/2024 16:06:20 [WARNING|DP=0|PP=5|TP=0|ip-26-0-169-207]: Repo card metadata block was not found. Setting CardData to empty. [default6]:Repo card metadata block was not found. Setting CardData to empty. [default2]:Repo card metadata block was not found. Setting CardData to empty. [default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default6]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default7]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default5]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default5]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default4]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default3]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default3]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default2]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default2]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: Attempting to run cuBLAS, but there was no current CUDA context! Attempting to set the primary context... (Triggered internally at ../aten/src/ATen/cuda/CublasHandlePool.cpp:135.) [default0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default1]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default6]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default7]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default4]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default5]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default5]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default2]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default2]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: Attempting to run cuBLAS, but there was no current CUDA context! Attempting to set the primary context... (Triggered internally at ../aten/src/ATen/cuda/CublasHandlePool.cpp:135.) [default0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default3]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default3]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.) [default1]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass [default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions [default7]: warnings.warn( [default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions [default1]: warnings.warn( [default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions [default6]: warnings.warn( [default0]:07/02/2024 16:07:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 2135.54MiB. Peak allocated 41452.48MiB. Peak reserved: 41696.00MiB [default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions [default0]: warnings.warn( [default0]:07/02/2024 16:07:21 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 4459.12MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:07:21 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 1 / 20 | consumed_tokens: 4.19M | elapsed_time_per_iteration_ms: 58.2K | tokens_per_sec: 72.1K | tokens_per_sec_per_gpu: 4.5K | global_batch_size: 1.02K | lm_loss: 11.1 | lr: 0.0001 | model_tflops_per_gpu: 40.9 | hardware_tflops_per_gpu: 40.9 | grad_norm: 24.9 | cuda_memory_allocated: 1.3G | cuda_max_memory_reserved: 12.8G | hd_total_memory_tb: 312G | hd_used_memory_tb: 65.5G | hd_free_memory_tb: 247G [default0]:07/02/2024 16:07:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:07:50 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 2 / 20 | consumed_tokens: 8.39M | elapsed_time_per_iteration_ms: 28.8K | tokens_per_sec: 146K | tokens_per_sec_per_gpu: 9.1K | global_batch_size: 1.02K | lm_loss: 11.1 | lr: 9.53e-05 | model_tflops_per_gpu: 82.5 | hardware_tflops_per_gpu: 82.5 | grad_norm: 25.1 | cuda_memory_allocated: 1.3G | cuda_max_memory_reserved: 12.8G | hd_total_memory_tb: 312G | hd_used_memory_tb: 65.5G | hd_free_memory_tb: 247G [default0]:07/02/2024 16:07:50 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 4459.12MiB. Peak reserved: 42994.00MiB [default0]:07/02/2024 16:08:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:08:18 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 3 / 20 | consumed_tokens: 12.6M | elapsed_time_per_iteration_ms: 28.7K | tokens_per_sec: 146K | tokens_per_sec_per_gpu: 9.13K | global_batch_size: 1.02K | lm_loss: 9.49 | lr: 9.05e-05 | model_tflops_per_gpu: 82.8 | hardware_tflops_per_gpu: 82.8 | grad_norm: 21.5 | cuda_memory_allocated: 1.3G | cuda_max_memory_reserved: 12.8G | hd_total_memory_tb: 312G | hd_used_memory_tb: 65.5G | hd_free_memory_tb: 247G [default0]:07/02/2024 16:08:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 4459.12MiB. Peak reserved: 42994.00MiB [default0]:STAGE:2024-07-02 16:08:18 687610:687610 ActivityProfilerController.cpp:314] Completed Stage: Warm Up [default0]:07/02/2024 16:08:47 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default0]:07/02/2024 16:08:48 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 4459.12MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:08:48 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 4 / 20 | consumed_tokens: 16.8M | elapsed_time_per_iteration_ms: 29.5K | tokens_per_sec: 142K | tokens_per_sec_per_gpu: 8.89K | global_batch_size: 1.02K | lm_loss: 9.36 | lr: 8.58e-05 | model_tflops_per_gpu: 80.6 | hardware_tflops_per_gpu: 80.6 | grad_norm: 21.4 | cuda_memory_allocated: 1.3G | cuda_max_memory_reserved: 12.8G | hd_total_memory_tb: 312G | hd_used_memory_tb: 65.5G | hd_free_memory_tb: 247G [default6]:07/02/2024 16:09:16 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 5 / 20 | consumed_tokens: 21M | elapsed_time_per_iteration_ms: 28.4K | tokens_per_sec: 148K | tokens_per_sec_per_gpu: 9.23K | global_batch_size: 1.02K | lm_loss: 9.02 | lr: 8.11e-05 | model_tflops_per_gpu: 83.8 | hardware_tflops_per_gpu: 83.8 | grad_norm: 12.7 [default0]:07/02/2024 16:09:16 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:09:47 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 6 / 20 | consumed_tokens: 25.2M | elapsed_time_per_iteration_ms: 30.8K | tokens_per_sec: 136K | tokens_per_sec_per_gpu: 8.51K | global_batch_size: 1.02K | lm_loss: 10.3 | lr: 7.63e-05 | model_tflops_per_gpu: 77.2 | hardware_tflops_per_gpu: 77.2 | grad_norm: 47.1 [default0]:STAGE:2024-07-02 16:10:00 687610:687610 ActivityProfilerController.cpp:320] Completed Stage: Collection [default0]:STAGE:2024-07-02 16:10:01 687610:687610 ActivityProfilerController.cpp:324] Completed Stage: Post Processing [default0]:07/02/2024 16:11:36 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default0]:07/02/2024 16:11:57 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:11:57 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 7 / 20 | consumed_tokens: 29.4M | elapsed_time_per_iteration_ms: 130K | tokens_per_sec: 32.2K | tokens_per_sec_per_gpu: 2.01K | global_batch_size: 1.02K | lm_loss: 8.68 | lr: 7.16e-05 | model_tflops_per_gpu: 18.2 | hardware_tflops_per_gpu: 18.2 | grad_norm: 5.58 [default0]:07/02/2024 16:12:25 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:12:25 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 8 / 20 | consumed_tokens: 33.6M | elapsed_time_per_iteration_ms: 27.4K | tokens_per_sec: 153K | tokens_per_sec_per_gpu: 9.55K | global_batch_size: 1.02K | lm_loss: 8.32 | lr: 6.68e-05 | model_tflops_per_gpu: 86.7 | hardware_tflops_per_gpu: 86.7 | grad_norm: 4.77 [default6]:07/02/2024 16:12:53 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 9 / 20 | consumed_tokens: 37.7M | elapsed_time_per_iteration_ms: 28.5K | tokens_per_sec: 147K | tokens_per_sec_per_gpu: 9.2K | global_batch_size: 1.02K | lm_loss: 7.95 | lr: 6.21e-05 | model_tflops_per_gpu: 83.5 | hardware_tflops_per_gpu: 83.5 | grad_norm: 3.31 [default0]:07/02/2024 16:12:53 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:13:20 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 10 / 20 | consumed_tokens: 41.9M | elapsed_time_per_iteration_ms: 26.4K | tokens_per_sec: 159K | tokens_per_sec_per_gpu: 9.94K | global_batch_size: 1.02K | lm_loss: 7.69 | lr: 5.74e-05 | model_tflops_per_gpu: 90.2 | hardware_tflops_per_gpu: 90.2 | grad_norm: 4.31 [default0]:07/02/2024 16:13:20 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default0]:07/02/2024 16:13:48 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:13:48 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 11 / 20 | consumed_tokens: 46.1M | elapsed_time_per_iteration_ms: 28.1K | tokens_per_sec: 149K | tokens_per_sec_per_gpu: 9.33K | global_batch_size: 1.02K | lm_loss: 7.45 | lr: 5.26e-05 | model_tflops_per_gpu: 84.7 | hardware_tflops_per_gpu: 84.7 | grad_norm: 2.5 [default6]:07/02/2024 16:14:18 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 12 / 20 | consumed_tokens: 50.3M | elapsed_time_per_iteration_ms: 30K | tokens_per_sec: 140K | tokens_per_sec_per_gpu: 8.73K | global_batch_size: 1.02K | lm_loss: 7.37 | lr: 4.79e-05 | model_tflops_per_gpu: 79.2 | hardware_tflops_per_gpu: 79.2 | grad_norm: 5.02 [default0]:07/02/2024 16:14:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:14:47 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 13 / 20 | consumed_tokens: 54.5M | elapsed_time_per_iteration_ms: 29.2K | tokens_per_sec: 144K | tokens_per_sec_per_gpu: 8.97K | global_batch_size: 1.02K | lm_loss: 7.31 | lr: 4.32e-05 | model_tflops_per_gpu: 81.4 | hardware_tflops_per_gpu: 81.4 | grad_norm: 6.03 [default0]:07/02/2024 16:14:47 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default0]:07/02/2024 16:15:16 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:15:16 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 14 / 20 | consumed_tokens: 58.7M | elapsed_time_per_iteration_ms: 28.5K | tokens_per_sec: 147K | tokens_per_sec_per_gpu: 9.2K | global_batch_size: 1.02K | lm_loss: 7.19 | lr: 3.84e-05 | model_tflops_per_gpu: 83.4 | hardware_tflops_per_gpu: 83.4 | grad_norm: 5.29 [default6]:07/02/2024 16:15:44 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 15 / 20 | consumed_tokens: 62.9M | elapsed_time_per_iteration_ms: 28.1K | tokens_per_sec: 149K | tokens_per_sec_per_gpu: 9.32K | global_batch_size: 1.02K | lm_loss: 7.06 | lr: 3.37e-05 | model_tflops_per_gpu: 84.6 | hardware_tflops_per_gpu: 84.6 | grad_norm: 2.7 [default0]:07/02/2024 16:15:44 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default0]:07/02/2024 16:16:12 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:16:12 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 16 / 20 | consumed_tokens: 67.1M | elapsed_time_per_iteration_ms: 28.4K | tokens_per_sec: 148K | tokens_per_sec_per_gpu: 9.25K | global_batch_size: 1.02K | lm_loss: 6.97 | lr: 2.89e-05 | model_tflops_per_gpu: 83.9 | hardware_tflops_per_gpu: 83.9 | grad_norm: 1.99 [default6]:07/02/2024 16:16:41 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 17 / 20 | consumed_tokens: 71.3M | elapsed_time_per_iteration_ms: 29.1K | tokens_per_sec: 144K | tokens_per_sec_per_gpu: 9.01K | global_batch_size: 1.02K | lm_loss: 6.91 | lr: 2.42e-05 | model_tflops_per_gpu: 81.7 | hardware_tflops_per_gpu: 81.7 | grad_norm: 2.01 [default0]:07/02/2024 16:16:41 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:17:09 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 18 / 20 | consumed_tokens: 75.5M | elapsed_time_per_iteration_ms: 28.2K | tokens_per_sec: 149K | tokens_per_sec_per_gpu: 9.29K | global_batch_size: 1.02K | lm_loss: 6.86 | lr: 1.95e-05 | model_tflops_per_gpu: 84.3 | hardware_tflops_per_gpu: 84.3 | grad_norm: 2.03 [default0]:07/02/2024 16:17:09 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:17:38 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 19 / 20 | consumed_tokens: 79.7M | elapsed_time_per_iteration_ms: 28.3K | tokens_per_sec: 148K | tokens_per_sec_per_gpu: 9.28K | global_batch_size: 1.02K | lm_loss: 6.81 | lr: 1.47e-05 | model_tflops_per_gpu: 84.2 | hardware_tflops_per_gpu: 84.2 | grad_norm: 2.04 [default0]:07/02/2024 16:17:38 [INFO|DP=0|PP=0|TP=0|ip-26-0-163-43]: Memory usage: 3168.25MiB. Peak allocated 42485.19MiB. Peak reserved: 42994.00MiB [default6]:07/02/2024 16:18:07 [INFO|DP=0|PP=7|TP=0|ip-26-0-169-207]: iteration: 20 / 20 | consumed_tokens: 83.9M | elapsed_time_per_iteration_ms: 29.1K | tokens_per_sec: 144K | tokens_per_sec_per_gpu: 9.02K | global_batch_size: 1.02K | lm_loss: 6.77 | lr: 1e-05 | model_tflops_per_gpu: 81.8 | hardware_tflops_per_gpu: 81.8 | grad_norm: 1.94 W0702 16:18:29.878000 139813752289024 torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1252] The node 'ip-26-0-169-207.ec2.internal_2256769_0' has failed to send a keep-alive heartbeat to the rendezvous 'none' due to an error of type RendezvousTimeoutError. Saved 1 csv files over 1 completed logs Processing file: /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-1_pp-8_mbz-4/profiler/ip-26-0-163-43_687610.1719936674744161495.pt.trace.json Results written to /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-1_pp-8_mbz-4/profiler.csv Consider using `hf_transfer` for faster uploads. This solution comes with some limitations. See https://huggingface.co/docs/huggingface_hub/hf_transfer for more details. ip-26-0-163-43_687610.1719936674744161495.pt.trace.json: 0%| | 0.00/3.06G [00:00