text
stringlengths
0
1.16k
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.attention.wo.lora_B.default.weight
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.feed_forward.w1.lora_A.default.weight
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.feed_forward.w1.lora_B.default.weight
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.feed_forward.w3.lora_A.default.weight
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.feed_forward.w3.lora_B.default.weight
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.feed_forward.w2.lora_A.default.weight
10/22/2024 17:07:54 - INFO - __main__ - language_model.base_model.model.model.layers.23.feed_forward.w2.lora_B.default.weight
[INFO|trainer.py:571] 2024-10-22 17:07:54,721 >> Using auto half precision backend
[2024-10-22 17:07:54,821] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed info: version=0.10.0, git-hash=unknown, git-branch=unknown
trainable params: 15,728,640 || all params: 1,904,875,520 || trainable%: 0.8257
trainable params: 15,728,640 || all params: 1,904,875,520 || trainable%: 0.8257
trainable params: 15,728,640 || all params: 1,904,875,520 || trainable%: 0.8257
[2024-10-22 17:07:56,459] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False
Using /mnt/SSD1_4TB/yunjie/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Using /mnt/SSD1_4TB/yunjie/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Using /mnt/SSD1_4TB/yunjie/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Using /mnt/SSD1_4TB/yunjie/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Detected CUDA files, patching ldflags
Emitting ninja build file /mnt/SSD1_4TB/yunjie/.cache/torch_extensions/py310_cu121/fused_adam/build.ninja...
Building extension module fused_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
ninja: no work to do.
Loading extension module fused_adam...
Time to load fused_adam op: 0.05200910568237305 seconds
Rank: 3 partition count [4] and sizes[(3932160, False)]
Loading extension module fused_adam...
Time to load fused_adam op: 0.10115861892700195 seconds
Loading extension module fused_adam...
Time to load fused_adam op: 0.10129737854003906 seconds
Loading extension module fused_adam...
Time to load fused_adam op: 0.10115528106689453 seconds
[2024-10-22 17:07:56,802] [INFO] [logging.py:96:log_dist] [Rank 0] Using DeepSpeed Optimizer param name adamw as basic optimizer
[2024-10-22 17:07:56,816] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Basic Optimizer = FusedAdam
[2024-10-22 17:07:56,816] [INFO] [utils.py:54:is_zero_supported_optimizer] Checking ZeRO support for optimizer=FusedAdam type=<class 'deepspeed.ops.adam.fused_adam.FusedAdam'>
[2024-10-22 17:07:56,816] [INFO] [logging.py:96:log_dist] [Rank 0] Creating torch.bfloat16 ZeRO stage 1 optimizer
[2024-10-22 17:07:56,816] [INFO] [stage_1_and_2.py:133:__init__] Reduce bucket size 1000000000
[2024-10-22 17:07:56,816] [INFO] [stage_1_and_2.py:134:__init__] Allgather bucket size 1000000000
[2024-10-22 17:07:56,816] [INFO] [stage_1_and_2.py:135:__init__] CPU Offload: False
[2024-10-22 17:07:56,816] [INFO] [stage_1_and_2.py:136:__init__] Round robin gradient partitioning: False
Rank: 2 partition count [4] and sizes[(3932160, False)]
Rank: 1 partition count [4] and sizes[(3932160, False)]
Rank: 0 partition count [4] and sizes[(3932160, False)]
[2024-10-22 17:07:57,265] [INFO] [utils.py:785:see_memory_usage] Before initializing optimizer states
[2024-10-22 17:07:57,266] [INFO] [utils.py:786:see_memory_usage] MA 4.94 GB Max_MA 4.95 GB CA 5.17 GB Max_CA 5 GB
[2024-10-22 17:07:57,266] [INFO] [utils.py:793:see_memory_usage] CPU Virtual Memory: used = 37.04 GB, percent = 7.4%
[2024-10-22 17:07:57,354] [INFO] [utils.py:785:see_memory_usage] After initializing optimizer states
[2024-10-22 17:07:57,355] [INFO] [utils.py:786:see_memory_usage] MA 4.97 GB Max_MA 4.99 GB CA 5.2 GB Max_CA 5 GB
[2024-10-22 17:07:57,355] [INFO] [utils.py:793:see_memory_usage] CPU Virtual Memory: used = 37.04 GB, percent = 7.4%
[2024-10-22 17:07:57,355] [INFO] [stage_1_and_2.py:493:__init__] optimizer state initialized
[2024-10-22 17:07:57,441] [INFO] [utils.py:785:see_memory_usage] After initializing ZeRO optimizer
[2024-10-22 17:07:57,442] [INFO] [utils.py:786:see_memory_usage] MA 4.97 GB Max_MA 4.97 GB CA 5.2 GB Max_CA 5 GB
[2024-10-22 17:07:57,442] [INFO] [utils.py:793:see_memory_usage] CPU Virtual Memory: used = 37.04 GB, percent = 7.4%
[2024-10-22 17:07:57,443] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Final Optimizer = adamw
[2024-10-22 17:07:57,444] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed using client callable to create LR scheduler
[2024-10-22 17:07:57,444] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed LR Scheduler = <torch.optim.lr_scheduler.LambdaLR object at 0x77ace29e0250>
[2024-10-22 17:07:57,444] [INFO] [logging.py:96:log_dist] [Rank 0] step=0, skipped=0, lr=[0.0], mom=[[0.9, 0.999]]
[2024-10-22 17:07:57,446] [INFO] [config.py:960:print] DeepSpeedEngine configuration:
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] activation_checkpointing_config {
"partition_activations": false,
"contiguous_memory_optimization": false,
"cpu_checkpointing": false,
"number_checkpoints": null,
"synchronize_checkpoint_boundary": false,
"profile": false
}
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'thread_count': 1, 'single_submit': False, 'overlap_events': True}
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] amp_enabled .................. False
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] amp_params ................... False
[2024-10-22 17:07:57,446] [INFO] [config.py:964: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
}
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] bfloat16_enabled ............. True
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] checkpoint_parallel_write_pipeline False
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] checkpoint_tag_validation_enabled True
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] checkpoint_tag_validation_fail False
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] comms_config ................. <deepspeed.comm.config.DeepSpeedCommsConfig object at 0x77acd1525b70>
[2024-10-22 17:07:57,446] [INFO] [config.py:964:print] communication_data_type ...... None
[2024-10-22 17:07:57,446] [INFO] [config.py:964: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}}