[2024-01-13 11:34:33,788] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2024-01-13 11:34:34,622] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. [2024-01-13 11:34:34,623] [INFO] [runner.py:571:main] cmd = /home/sp-operator/hf_env/bin/python3 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMF19 --master_addr=127.0.0.1 --master_port=29500 --enable_each_rank_log=None /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/run_speech_recognition_seq2seq_streaming.py --deepspeed=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/ds_config.json --model_name_or_path=openai/whisper-tiny --dataset_name=mozilla-foundation/common_voice_16_0 --dataset_config_name=id --language=Indonesian --train_split_name=train --eval_split_name=test --model_index_name=Breeze DSW Indonesian - tiny --max_steps=2000 --output_dir=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id --per_device_train_batch_size=32 --per_device_eval_batch_size=16 --logging_steps=25 --learning_rate=1e-5 --warmup_steps=500 --evaluation_strategy=steps --eval_steps=200 --save_strategy=steps --save_steps=200 --generation_max_length=225 --length_column_name=input_length --max_duration_in_seconds=30 --text_column_name=sentence --freeze_feature_encoder=False --report_to=tensorboard --metric_for_best_model=wer --greater_is_better=False --load_best_model_at_end --gradient_checkpointing --fp16 --do_train --do_eval --predict_with_generate --do_normalize_eval --streaming --use_auth_token --push_to_hub [2024-01-13 11:34:36,328] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2024-01-13 11:34:37,085] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0]} [2024-01-13 11:34:37,085] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=1, node_rank=0 [2024-01-13 11:34:37,085] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(, {'localhost': [0]}) [2024-01-13 11:34:37,085] [INFO] [launch.py:163:main] dist_world_size=1 [2024-01-13 11:34:37,085] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0 [2024-01-13 11:34:40,302] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2024-01-13 11:34:40,485] [INFO] [comm.py:637:init_distributed] cdb=None [2024-01-13 11:34:40,485] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl 01/13/2024 11:34:40 - WARNING - __main__ - Process rank: 0, device: cuda:0, n_gpu: 1distributed training: True, 16-bits training: True 01/13/2024 11:34:40 - INFO - __main__ - Training/evaluation parameters Seq2SeqTrainingArguments( _n_gpu=1, adafactor=False, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, auto_find_batch_size=False, bf16=False, bf16_full_eval=False, data_seed=None, dataloader_drop_last=False, dataloader_num_workers=0, dataloader_persistent_workers=False, dataloader_pin_memory=True, ddp_backend=None, ddp_broadcast_buffers=None, ddp_bucket_cap_mb=None, ddp_find_unused_parameters=None, ddp_timeout=1800, debug=[], deepspeed=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/ds_config.json, disable_tqdm=False, dispatch_batches=None, do_eval=True, do_predict=False, do_train=True, eval_accumulation_steps=None, eval_delay=0, eval_steps=200, evaluation_strategy=steps, fp16=True, fp16_backend=auto, fp16_full_eval=False, fp16_opt_level=O1, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_grad_ckpt': False}, fsdp_min_num_params=0, fsdp_transformer_layer_cls_to_wrap=None, full_determinism=False, generation_config=None, generation_max_length=225, generation_num_beams=None, gradient_accumulation_steps=1, gradient_checkpointing=True, gradient_checkpointing_kwargs=None, greater_is_better=False, group_by_length=False, half_precision_backend=auto, hub_always_push=False, hub_model_id=None, hub_private_repo=False, hub_strategy=every_save, hub_token=, ignore_data_skip=False, include_inputs_for_metrics=False, include_num_input_tokens_seen=False, include_tokens_per_second=False, jit_mode_eval=False, label_names=None, label_smoothing_factor=0.0, learning_rate=1e-05, length_column_name=input_length, load_best_model_at_end=True, local_rank=0, log_level=passive, log_level_replica=warning, log_on_each_node=True, logging_dir=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/runs/Jan13_11-34-39_knight, logging_first_step=False, logging_nan_inf_filter=True, logging_steps=25, logging_strategy=steps, lr_scheduler_kwargs={}, lr_scheduler_type=linear, max_grad_norm=1.0, max_steps=2000, metric_for_best_model=wer, mp_parameters=, neftune_noise_alpha=None, no_cuda=False, num_train_epochs=3.0, optim=adamw_torch, optim_args=None, output_dir=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id, overwrite_output_dir=False, past_index=-1, per_device_eval_batch_size=16, per_device_train_batch_size=32, predict_with_generate=True, prediction_loss_only=False, push_to_hub=True, push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=, ray_scope=last, remove_unused_columns=True, report_to=['tensorboard'], resume_from_checkpoint=None, run_name=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id, save_on_each_node=False, save_only_model=False, save_safetensors=True, save_steps=200, save_strategy=steps, save_total_limit=None, seed=42, skip_memory_metrics=True, sortish_sampler=False, split_batches=False, tf32=None, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, torchdynamo=None, tpu_metrics_debug=False, tpu_num_cores=None, use_cpu=False, use_ipex=False, use_legacy_prediction_loop=False, use_mps_device=False, warmup_ratio=0.0, warmup_steps=500, weight_decay=0.0, ) 01/13/2024 11:34:40 - INFO - __main__ - Training/evaluation parameters Seq2SeqTrainingArguments( _n_gpu=1, adafactor=False, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, auto_find_batch_size=False, bf16=False, bf16_full_eval=False, data_seed=None, dataloader_drop_last=False, dataloader_num_workers=0, dataloader_persistent_workers=False, dataloader_pin_memory=True, ddp_backend=None, ddp_broadcast_buffers=None, ddp_bucket_cap_mb=None, ddp_find_unused_parameters=None, ddp_timeout=1800, debug=[], deepspeed=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/ds_config.json, disable_tqdm=False, dispatch_batches=None, do_eval=True, do_predict=False, do_train=True, eval_accumulation_steps=None, eval_delay=0, eval_steps=200, evaluation_strategy=steps, fp16=True, fp16_backend=auto, fp16_full_eval=False, fp16_opt_level=O1, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_grad_ckpt': False}, fsdp_min_num_params=0, fsdp_transformer_layer_cls_to_wrap=None, full_determinism=False, generation_config=None, generation_max_length=225, generation_num_beams=None, gradient_accumulation_steps=1, gradient_checkpointing=True, gradient_checkpointing_kwargs=None, greater_is_better=False, group_by_length=False, half_precision_backend=auto, hub_always_push=False, hub_model_id=None, hub_private_repo=False, hub_strategy=every_save, hub_token=, ignore_data_skip=False, include_inputs_for_metrics=False, include_num_input_tokens_seen=False, include_tokens_per_second=False, jit_mode_eval=False, label_names=None, label_smoothing_factor=0.0, learning_rate=1e-05, length_column_name=input_length, load_best_model_at_end=True, local_rank=0, log_level=passive, log_level_replica=warning, log_on_each_node=True, logging_dir=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/runs/Jan13_11-34-39_knight, logging_first_step=False, logging_nan_inf_filter=True, logging_steps=25, logging_strategy=steps, lr_scheduler_kwargs={}, lr_scheduler_type=linear, max_grad_norm=1.0, max_steps=2000, metric_for_best_model=wer, mp_parameters=, neftune_noise_alpha=None, no_cuda=False, num_train_epochs=3.0, optim=adamw_torch, optim_args=None, output_dir=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id, overwrite_output_dir=False, past_index=-1, per_device_eval_batch_size=16, per_device_train_batch_size=32, predict_with_generate=True, prediction_loss_only=False, push_to_hub=True, push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=, ray_scope=last, remove_unused_columns=True, report_to=['tensorboard'], resume_from_checkpoint=None, run_name=/cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id, save_on_each_node=False, save_only_model=False, save_safetensors=True, save_steps=200, save_strategy=steps, save_total_limit=None, seed=42, skip_memory_metrics=True, sortish_sampler=False, split_batches=False, tf32=None, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, torchdynamo=None, tpu_metrics_debug=False, tpu_num_cores=None, use_cpu=False, use_ipex=False, use_legacy_prediction_loop=False, use_mps_device=False, warmup_ratio=0.0, warmup_steps=500, weight_decay=0.0, ) 01/13/2024 11:34:40 - INFO - __main__ - Checkpoint detected, resuming training at /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500. To avoid this behavior, change the `--output_dir` or add `--overwrite_output_dir` to train from scratch. [2024-01-13 11:34:56,451] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed info: version=0.12.6, git-hash=unknown, git-branch=unknown [2024-01-13 11:34:57,219] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False Installed CUDA version 12.3 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination [1/4] /usr/local/cuda/bin/nvcc -DTORCH_EXTENSION_NAME=cpu_adam -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -I/home/sp-operator/hf_env/lib/python3.10/site-packages/deepspeed/ops/csrc/includes -I/usr/local/cuda/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/TH -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /usr/include/python3.10 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_61,code=compute_61 -gencode=arch=compute_61,code=sm_61 --compiler-options '-fPIC' -O3 --use_fast_math -std=c++17 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ -U__CUDA_NO_HALF2_OPERATORS__ -gencode=arch=compute_61,code=sm_61 -gencode=arch=compute_61,code=compute_61 -c /home/sp-operator/hf_env/lib/python3.10/site-packages/deepspeed/ops/csrc/common/custom_cuda_kernel.cu -o custom_cuda_kernel.cuda.o [2/4] c++ -MMD -MF cpu_adam.o.d -DTORCH_EXTENSION_NAME=cpu_adam -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -I/home/sp-operator/hf_env/lib/python3.10/site-packages/deepspeed/ops/csrc/includes -I/usr/local/cuda/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/TH -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /usr/include/python3.10 -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++17 -O3 -std=c++17 -g -Wno-reorder -L/usr/local/cuda/lib64 -lcudart -lcublas -g -march=native -fopenmp -D__AVX256__ -D__ENABLE_CUDA__ -c /home/sp-operator/hf_env/lib/python3.10/site-packages/deepspeed/ops/csrc/adam/cpu_adam.cpp -o cpu_adam.o [3/4] c++ -MMD -MF cpu_adam_impl.o.d -DTORCH_EXTENSION_NAME=cpu_adam -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -I/home/sp-operator/hf_env/lib/python3.10/site-packages/deepspeed/ops/csrc/includes -I/usr/local/cuda/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/TH -isystem /home/sp-operator/hf_env/lib/python3.10/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /usr/include/python3.10 -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++17 -O3 -std=c++17 -g -Wno-reorder -L/usr/local/cuda/lib64 -lcudart -lcublas -g -march=native -fopenmp -D__AVX256__ -D__ENABLE_CUDA__ -c /home/sp-operator/hf_env/lib/python3.10/site-packages/deepspeed/ops/csrc/adam/cpu_adam_impl.cpp -o cpu_adam_impl.o [4/4] c++ cpu_adam.o cpu_adam_impl.o custom_cuda_kernel.cuda.o -shared -lcurand -L/home/sp-operator/hf_env/lib/python3.10/site-packages/torch/lib -lc10 -lc10_cuda -ltorch_cpu -ltorch_cuda -ltorch -ltorch_python -L/usr/local/cuda/lib64 -lcudart -o cpu_adam.so Time to load cpu_adam op: 26.69891905784607 seconds Adam Optimizer #0 is created with AVX2 arithmetic capability. Config: alpha=0.000010, betas=(0.900000, 0.999000), weight_decay=0.000000, adam_w=1 [2024-01-13 11:35:25,684] [INFO] [logging.py:96:log_dist] [Rank 0] Using DeepSpeed Optimizer param name adamw as basic optimizer [2024-01-13 11:35:25,684] [INFO] [logging.py:96:log_dist] [Rank 0] Removing param_group that has no 'params' in the basic Optimizer [2024-01-13 11:35:25,689] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Basic Optimizer = DeepSpeedCPUAdam [2024-01-13 11:35:25,689] [INFO] [utils.py:56:is_zero_supported_optimizer] Checking ZeRO support for optimizer=DeepSpeedCPUAdam type= [2024-01-13 11:35:25,689] [INFO] [logging.py:96:log_dist] [Rank 0] Creating torch.float16 ZeRO stage 2 optimizer [2024-01-13 11:35:25,689] [INFO] [stage_1_and_2.py:148:__init__] Reduce bucket size 200000000 [2024-01-13 11:35:25,689] [INFO] [stage_1_and_2.py:149:__init__] Allgather bucket size 200000000 [2024-01-13 11:35:25,689] [INFO] [stage_1_and_2.py:150:__init__] CPU Offload: True [2024-01-13 11:35:25,689] [INFO] [stage_1_and_2.py:151:__init__] Round robin gradient partitioning: False [2024-01-13 11:35:25,955] [INFO] [utils.py:791:see_memory_usage] Before initializing optimizer states [2024-01-13 11:35:25,956] [INFO] [utils.py:792:see_memory_usage] MA 0.11 GB Max_MA 0.11 GB CA 0.13 GB Max_CA 0 GB [2024-01-13 11:35:25,956] [INFO] [utils.py:799:see_memory_usage] CPU Virtual Memory: used = 3.6 GB, percent = 5.7% [2024-01-13 11:35:26,212] [INFO] [utils.py:791:see_memory_usage] After initializing optimizer states [2024-01-13 11:35:26,213] [INFO] [utils.py:792:see_memory_usage] MA 0.11 GB Max_MA 0.11 GB CA 0.13 GB Max_CA 0 GB [2024-01-13 11:35:26,213] [INFO] [utils.py:799:see_memory_usage] CPU Virtual Memory: used = 4.12 GB, percent = 6.6% [2024-01-13 11:35:26,213] [INFO] [stage_1_and_2.py:516:__init__] optimizer state initialized [2024-01-13 11:35:26,292] [INFO] [utils.py:791:see_memory_usage] After initializing ZeRO optimizer [2024-01-13 11:35:26,292] [INFO] [utils.py:792:see_memory_usage] MA 0.11 GB Max_MA 0.11 GB CA 0.13 GB Max_CA 0 GB [2024-01-13 11:35:26,292] [INFO] [utils.py:799:see_memory_usage] CPU Virtual Memory: used = 4.12 GB, percent = 6.6% [2024-01-13 11:35:26,301] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Final Optimizer = adamw [2024-01-13 11:35:26,301] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed using configured LR scheduler = WarmupDecayLR [2024-01-13 11:35:26,301] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed LR Scheduler = [2024-01-13 11:35:26,301] [INFO] [logging.py:96:log_dist] [Rank 0] step=0, skipped=0, lr=[1e-05], mom=[[0.9, 0.999]] [2024-01-13 11:35:26,301] [INFO] [config.py:984:print] DeepSpeedEngine configuration: [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] activation_checkpointing_config { "partition_activations": false, "contiguous_memory_optimization": false, "cpu_checkpointing": false, "number_checkpoints": null, "synchronize_checkpoint_boundary": false, "profile": false } [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'thread_count': 1, 'single_submit': False, 'overlap_events': True} [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] amp_enabled .................. False [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] amp_params ................... False [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] autotuning_config ............ { "enabled": false, "start_step": null, "end_step": null, "metric_path": null, "arg_mappings": null, "metric": "throughput", "model_info": null, "results_dir": "autotuning_results", "exps_dir": "autotuning_exps", "overwrite": true, "fast": true, "start_profile_step": 3, "end_profile_step": 5, "tuner_type": "gridsearch", "tuner_early_stopping": 5, "tuner_num_trials": 50, "model_info_path": null, "mp_size": 1, "max_train_batch_size": null, "min_train_batch_size": 1, "max_train_micro_batch_size_per_gpu": 1.024000e+03, "min_train_micro_batch_size_per_gpu": 1, "num_tuning_micro_batch_sizes": 3 } [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] bfloat16_enabled ............. False [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] checkpoint_parallel_write_pipeline False [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] checkpoint_tag_validation_enabled True [2024-01-13 11:35:26,302] [INFO] [config.py:988:print] checkpoint_tag_validation_fail False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] comms_config ................. [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] communication_data_type ...... None [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] compression_config ........... {'weight_quantization': {'shared_parameters': {'enabled': False, 'quantizer_kernel': False, 'schedule_offset': 0, 'quantize_groups': 1, 'quantize_verbose': False, 'quantization_type': 'symmetric', 'quantize_weight_in_forward': False, 'rounding': 'nearest', 'fp16_mixed_quantize': False, 'quantize_change_ratio': 0.001}, 'different_groups': {}}, 'activation_quantization': {'shared_parameters': {'enabled': False, 'quantization_type': 'symmetric', 'range_calibration': 'dynamic', 'schedule_offset': 1000}, 'different_groups': {}}, 'sparse_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'row_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'head_pruning': {'shared_parameters': {'enabled': False, 'method': 'topk', 'schedule_offset': 1000}, 'different_groups': {}}, 'channel_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'layer_reduction': {'enabled': False}} [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] curriculum_enabled_legacy .... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] curriculum_params_legacy ..... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] data_efficiency_config ....... {'enabled': False, 'seed': 1234, 'data_sampling': {'enabled': False, 'num_epochs': 1000, 'num_workers': 0, 'curriculum_learning': {'enabled': False}}, 'data_routing': {'enabled': False, 'random_ltd': {'enabled': False, 'layer_token_lr_schedule': {'enabled': False}}}} [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] data_efficiency_enabled ...... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] dataloader_drop_last ......... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] disable_allgather ............ False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] dump_state ................... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] dynamic_loss_scale_args ...... {'init_scale': 65536, 'scale_window': 1000, 'delayed_shift': 2, 'consecutive_hysteresis': False, 'min_scale': 1} [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_enabled ........... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_gas_boundary_resolution 1 [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_layer_name ........ bert.encoder.layer [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_layer_num ......... 0 [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_max_iter .......... 100 [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_stability ......... 1e-06 [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_tol ............... 0.01 [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] eigenvalue_verbose ........... False [2024-01-13 11:35:26,303] [INFO] [config.py:988:print] elasticity_enabled ........... False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] flops_profiler_config ........ { "enabled": false, "recompute_fwd_factor": 0.0, "profile_step": 1, "module_depth": -1, "top_modules": 1, "detailed": true, "output_file": null } [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] fp16_auto_cast ............... False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] fp16_enabled ................. True [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] fp16_master_weights_and_gradients False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] global_rank .................. 0 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] grad_accum_dtype ............. None [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] gradient_accumulation_steps .. 1 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] gradient_clipping ............ 1.0 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] gradient_predivide_factor .... 1.0 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] graph_harvesting ............. False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] hybrid_engine ................ enabled=False max_out_tokens=512 inference_tp_size=1 release_inference_cache=False pin_parameters=True tp_gather_partition_size=8 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] initial_dynamic_scale ........ 65536 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] load_universal_checkpoint .... False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] loss_scale ................... 0 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] memory_breakdown ............. False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] mics_hierarchial_params_gather False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] mics_shard_size .............. -1 [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] monitor_config ............... tensorboard=TensorBoardConfig(enabled=False, output_path='', job_name='DeepSpeedJobName') wandb=WandbConfig(enabled=False, group=None, team=None, project='deepspeed') csv_monitor=CSVConfig(enabled=False, output_path='', job_name='DeepSpeedJobName') enabled=False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] nebula_config ................ { "enabled": false, "persistent_storage_path": null, "persistent_time_interval": 100, "num_of_version_in_retention": 2, "enable_nebula_load": true, "load_path": null } [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] optimizer_legacy_fusion ...... False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] optimizer_name ............... adamw [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] optimizer_params ............. {'lr': 1e-05, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0} [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] pipeline ..................... {'stages': 'auto', 'partition': 'best', 'seed_layers': False, 'activation_checkpoint_interval': 0, 'pipe_partitioned': True, 'grad_partitioned': True} [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] pld_enabled .................. False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] pld_params ................... False [2024-01-13 11:35:26,304] [INFO] [config.py:988:print] prescale_gradients ........... False [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] scheduler_name ............... WarmupDecayLR [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] scheduler_params ............. {'last_batch_iteration': -1, 'total_num_steps': 2000, 'warmup_min_lr': 0, 'warmup_max_lr': 1e-05, 'warmup_num_steps': 500} [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] seq_parallel_communication_data_type torch.float32 [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] sparse_attention ............. None [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] sparse_gradients_enabled ..... False [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] steps_per_print .............. inf [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] train_batch_size ............. 32 [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] train_micro_batch_size_per_gpu 32 [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] use_data_before_expert_parallel_ False [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] use_node_local_storage ....... False [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] wall_clock_breakdown ......... False [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] weight_quantization_config ... None [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] world_size ................... 1 [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] zero_allow_untested_optimizer False [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] zero_config .................. stage=2 contiguous_gradients=True reduce_scatter=True reduce_bucket_size=200000000 use_multi_rank_bucket_allreduce=True allgather_partitions=True allgather_bucket_size=200000000 overlap_comm=True load_from_fp32_weights=True elastic_checkpoint=False offload_param=None offload_optimizer=DeepSpeedZeroOffloadOptimizerConfig(device='cpu', nvme_path=None, buffer_count=4, pin_memory=True, pipeline=False, pipeline_read=False, pipeline_write=False, fast_init=False, ratio=1.0) sub_group_size=1,000,000,000 cpu_offload_param=None cpu_offload_use_pin_memory=None cpu_offload=None prefetch_bucket_size=50,000,000 param_persistence_threshold=100,000 model_persistence_threshold=sys.maxsize max_live_parameters=1,000,000,000 max_reuse_distance=1,000,000,000 gather_16bit_weights_on_model_save=False stage3_gather_fp16_weights_on_model_save=False ignore_unused_parameters=True legacy_stage1=False round_robin_gradients=False zero_hpz_partition_size=1 zero_quantized_weights=False zero_quantized_nontrainable_weights=False zero_quantized_gradients=False mics_shard_size=-1 mics_hierarchical_params_gather=False memory_efficient_linear=True pipeline_loading_checkpoint=False override_module_apply=True [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] zero_enabled ................. True [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] zero_force_ds_cpu_optimizer .. True [2024-01-13 11:35:26,305] [INFO] [config.py:988:print] zero_optimization_stage ...... 2 [2024-01-13 11:35:26,305] [INFO] [config.py:974:print_user_config] json = { "fp16": { "enabled": true, "loss_scale": 0, "loss_scale_window": 1000, "initial_scale_power": 16, "hysteresis": 2, "min_loss_scale": 1 }, "optimizer": { "type": "AdamW", "params": { "lr": 1e-05, "betas": [0.9, 0.999], "eps": 1e-08, "weight_decay": 0.0 } }, "scheduler": { "type": "WarmupDecayLR", "params": { "last_batch_iteration": -1, "total_num_steps": 2.000000e+03, "warmup_min_lr": 0, "warmup_max_lr": 1e-05, "warmup_num_steps": 500 } }, "zero_optimization": { "stage": 2, "offload_optimizer": { "device": "cpu", "pin_memory": true }, "allgather_partitions": true, "allgather_bucket_size": 2.000000e+08, "overlap_comm": true, "reduce_scatter": true, "reduce_bucket_size": 2.000000e+08, "contiguous_gradients": true }, "gradient_accumulation_steps": 1, "gradient_clipping": 1.0, "train_batch_size": 32, "train_micro_batch_size_per_gpu": 32, "steps_per_print": inf, "bf16": { "enabled": false } } [2024-01-13 11:35:26,322] [INFO] [torch_checkpoint_engine.py:27:load] [Torch] Loading checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt... [2024-01-13 11:35:27,240] [INFO] [torch_checkpoint_engine.py:29:load] [Torch] Loaded checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt. [2024-01-13 11:35:27,249] [INFO] [torch_checkpoint_engine.py:27:load] [Torch] Loading checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt... [2024-01-13 11:35:27,294] [INFO] [torch_checkpoint_engine.py:29:load] [Torch] Loaded checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt. [2024-01-13 11:35:27,354] [INFO] [torch_checkpoint_engine.py:27:load] [Torch] Loading checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_optim_states.pt... [2024-01-13 11:35:32,554] [INFO] [torch_checkpoint_engine.py:29:load] [Torch] Loaded checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_optim_states.pt. [2024-01-13 11:35:32,554] [INFO] [engine.py:2988:_get_all_zero_checkpoint_state_dicts] successfully read 1 ZeRO state_dicts for rank 0 [2024-01-13 11:35:32,588] [INFO] [engine.py:2920:_load_zero_checkpoint] loading 1 zero partition checkpoints for rank 0 [2024-01-13 11:36:34,189] [INFO] [torch_checkpoint_engine.py:27:load] [Torch] Loading checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt... [2024-01-13 11:36:35,491] [INFO] [torch_checkpoint_engine.py:29:load] [Torch] Loaded checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt. [2024-01-13 11:36:35,509] [INFO] [torch_checkpoint_engine.py:27:load] [Torch] Loading checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt... [2024-01-13 11:36:35,554] [INFO] [torch_checkpoint_engine.py:29:load] [Torch] Loaded checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/mp_rank_00_model_states.pt. [2024-01-13 11:36:35,636] [INFO] [torch_checkpoint_engine.py:27:load] [Torch] Loading checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_optim_states.pt... [2024-01-13 11:36:40,068] [INFO] [torch_checkpoint_engine.py:29:load] [Torch] Loaded checkpoint from /cosmos/home/sp-operator/ai/training/models/huggingface/scripts/../breeze-dsw-tiny-id/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_optim_states.pt. [2024-01-13 11:36:40,068] [INFO] [engine.py:2988:_get_all_zero_checkpoint_state_dicts] successfully read 1 ZeRO state_dicts for rank 0 [2024-01-13 11:36:40,102] [INFO] [engine.py:2920:_load_zero_checkpoint] loading 1 zero partition checkpoints for rank 0 {'train_runtime': 67.5132, 'train_samples_per_second': 947.963, 'train_steps_per_second': 29.624, 'train_loss': 0.0, 'epoch': 3.03} ***** train metrics ***** epoch = 3.03 train_loss = 0.0 train_runtime = 0:01:07.51 train_samples_per_second = 947.963 train_steps_per_second = 29.624 01/13/2024 11:36:47 - INFO - __main__ - *** Evaluate *** ***** eval metrics ***** epoch = 3.03 eval_loss = 0.7109 eval_runtime = 0:18:47.23 eval_samples_per_second = 3.231 eval_steps_per_second = 0.202 eval_wer = 45.2434