|
[2023-12-11 05:39:03,031] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) |
|
[2023-12-11 05:39:04,827] [WARNING] [runner.py:203:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. |
|
[2023-12-11 05:39:04,828] [INFO] [runner.py:570:main] cmd = /home/t-sokumar/miniconda3/envs/ft/bin/python -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgM119 --master_addr=127.0.0.1 --master_port=29500 --enable_each_rank_log=None main.py --data_path local/jsonfile --data_split 1,0,0 --model_name_or_path codellama/CodeLlama-7b-hf --per_device_train_batch_size 8 --per_device_eval_batch_size 8 --max_seq_len 512 --learning_rate 9.65e-6 --weight_decay 0. --num_train_epochs 3 --gradient_accumulation_steps 1 --lr_scheduler_type cosine --num_warmup_steps 0 --seed 1234 --gradient_checkpointing --zero_stage 3 --deepspeed --lora_dim 128 --lora_module_name layers. --output_dir ./output_step1_Codellama_7b_lora_llamahub-devrev --add_eot_token |
|
[2023-12-11 05:39:07,364] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) |
|
[2023-12-11 05:39:09,159] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3]} |
|
[2023-12-11 05:39:09,159] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=4, node_rank=0 |
|
[2023-12-11 05:39:09,159] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(<class 'list'>, {'localhost': [0, 1, 2, 3]}) |
|
[2023-12-11 05:39:09,159] [INFO] [launch.py:163:main] dist_world_size=4 |
|
[2023-12-11 05:39:09,159] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3 |
|
[2023-12-11 05:39:12,594] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) |
|
[2023-12-11 05:39:12,600] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) |
|
[2023-12-11 05:39:12,605] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) |
|
[2023-12-11 05:39:12,606] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/transformers/deepspeed.py:23: FutureWarning: transformers.deepspeed module is deprecated and will be removed in a future version. Please import deepspeed modules directly from transformers.integrations |
|
warnings.warn( |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/transformers/deepspeed.py:23: FutureWarning: transformers.deepspeed module is deprecated and will be removed in a future version. Please import deepspeed modules directly from transformers.integrations |
|
warnings.warn( |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/transformers/deepspeed.py:23: FutureWarning: transformers.deepspeed module is deprecated and will be removed in a future version. Please import deepspeed modules directly from transformers.integrations |
|
warnings.warn( |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/transformers/deepspeed.py:23: FutureWarning: transformers.deepspeed module is deprecated and will be removed in a future version. Please import deepspeed modules directly from transformers.integrations |
|
warnings.warn( |
|
[2023-12-11 05:39:14,179] [INFO] [comm.py:637:init_distributed] cdb=None |
|
[2023-12-11 05:39:14,179] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl |
|
[2023-12-11 05:39:14,642] [INFO] [comm.py:637:init_distributed] cdb=None |
|
[2023-12-11 05:39:14,646] [INFO] [comm.py:637:init_distributed] cdb=None |
|
[2023-12-11 05:39:14,678] [INFO] [comm.py:637:init_distributed] cdb=None |
|
The tokenizer class you load from this checkpoint is not the same type as the class this function is called from. It may result in unexpected tokenization. |
|
The tokenizer class you load from this checkpoint is 'CodeLlamaTokenizer'. |
|
The class this function is called from is 'LlamaTokenizer'. |
|
The tokenizer class you load from this checkpoint is not the same type as the class this function is called from. It may result in unexpected tokenization. |
|
The tokenizer class you load from this checkpoint is 'CodeLlamaTokenizer'. |
|
The class this function is called from is 'LlamaTokenizer'. |
|
The tokenizer class you load from this checkpoint is not the same type as the class this function is called from. It may result in unexpected tokenization. |
|
The tokenizer class you load from this checkpoint is 'CodeLlamaTokenizer'. |
|
The class this function is called from is 'LlamaTokenizer'. |
|
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama.LlamaTokenizer'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thouroughly read the reason why this was added as explained in https: |
|
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama.LlamaTokenizer'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thouroughly read the reason why this was added as explained in https: |
|
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama.LlamaTokenizer'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thouroughly read the reason why this was added as explained in https: |
|
The tokenizer class you load from this checkpoint is not the same type as the class this function is called from. It may result in unexpected tokenization. |
|
The tokenizer class you load from this checkpoint is 'CodeLlamaTokenizer'. |
|
The class this function is called from is 'LlamaTokenizer'. |
|
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama.LlamaTokenizer'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thouroughly read the reason why this was added as explained in https: |
|
[2023-12-11 05:39:17,564] [INFO] [partition_parameters.py:348:__exit__] finished initializing model - num_params = 291, num_elems = 6.74B |
|
Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards: 50%|███████████████████████████████████████████████████████████▌ | 1/2 [00:04<00:04, 4.90s/it]
Loading checkpoint shards: 50%|███████████████████████████████████████████████████████████▌ | 1/2 [00:04<00:04, 4.91s/it]
Loading checkpoint shards: 50%|███████████████████████████████████████████████████████████▌ | 1/2 [00:04<00:04, 4.90s/it]
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 6.00s/it]
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 5.84s/it] |
|
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 6.00s/it]
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 5.84s/it] |
|
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 6.00s/it]
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:11<00:00, 5.84s/it] |
|
Loading checkpoint shards: 50%|███████████████████████████████████████████████████████████▌ | 1/2 [00:20<00:20, 20.85s/it]
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:28<00:00, 12.90s/it]
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:28<00:00, 14.09s/it] |
|
Using /home/t-sokumar/.cache/torch_extensions/py311_cu121 as PyTorch extensions root... |
|
Using /home/t-sokumar/.cache/torch_extensions/py311_cu121 as PyTorch extensions root... |
|
Using /home/t-sokumar/.cache/torch_extensions/py311_cu121 as PyTorch extensions root... |
|
Using /home/t-sokumar/.cache/torch_extensions/py311_cu121 as PyTorch extensions root... |
|
Detected CUDA files, patching ldflags |
|
Emitting ninja build file /home/t-sokumar/.cache/torch_extensions/py311_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: 0.12226700782775879 seconds |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/deepspeed/ops/adam/fused_adam.py:96: UserWarning: The torch.cuda.*DtypeTensor constructors are no longer recommended. It's best to use methods such as torch.tensor(data, dtype=*, device='cuda') to create tensors. (Triggered internally at ../torch/csrc/tensor/python_tensor.cpp:83.) |
|
self._dummy_overflow_buf = get_accelerator().IntTensor([0]) |
|
Loading extension module fused_adam...Loading extension module fused_adam... |
|
|
|
Time to load fused_adam op: 0.2073993682861328 secondsTime to load fused_adam op: 0.2075939178466797 seconds |
|
|
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/deepspeed/ops/adam/fused_adam.py:96: UserWarning: The torch.cuda.*DtypeTensor constructors are no longer recommended. It's best to use methods such as torch.tensor(data, dtype=*, device='cuda') to create tensors. (Triggered internally at ../torch/csrc/tensor/python_tensor.cpp:83.) |
|
self._dummy_overflow_buf = get_accelerator().IntTensor([0]) |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/deepspeed/ops/adam/fused_adam.py:96: UserWarning: The torch.cuda.*DtypeTensor constructors are no longer recommended. It's best to use methods such as torch.tensor(data, dtype=*, device='cuda') to create tensors. (Triggered internally at ../torch/csrc/tensor/python_tensor.cpp:83.) |
|
self._dummy_overflow_buf = get_accelerator().IntTensor([0]) |
|
Loading extension module fused_adam... |
|
[2023-12-11 05:39:51,139] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed info: version=0.12.4, git-hash=unknown, git-branch=unknown |
|
[2023-12-11 05:39:51,139] [INFO] [comm.py:662:init_distributed] Distributed backend already initialized |
|
Time to load fused_adam op: 0.2015979290008545 seconds |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/deepspeed/ops/adam/fused_adam.py:96: UserWarning: The torch.cuda.*DtypeTensor constructors are no longer recommended. It's best to use methods such as torch.tensor(data, dtype=*, device='cuda') to create tensors. (Triggered internally at ../torch/csrc/tensor/python_tensor.cpp:83.) |
|
self._dummy_overflow_buf = get_accelerator().IntTensor([0]) |
|
[2023-12-11 05:39:51,163] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False |
|
[2023-12-11 05:39:51,164] [INFO] [logging.py:96:log_dist] [Rank 0] Using client Optimizer as basic optimizer |
|
[2023-12-11 05:39:51,164] [INFO] [logging.py:96:log_dist] [Rank 0] Removing param_group that has no 'params' in the basic Optimizer |
|
[2023-12-11 05:39:51,206] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Basic Optimizer = FusedAdam |
|
[2023-12-11 05:39:51,206] [INFO] [utils.py:56:is_zero_supported_optimizer] Checking ZeRO support for optimizer=FusedAdam type=<class 'deepspeed.ops.adam.fused_adam.FusedAdam'> |
|
[2023-12-11 05:39:51,206] [INFO] [logging.py:96:log_dist] [Rank 0] Creating fp16 ZeRO stage 3 optimizer, MiCS is enabled False, Hierarchical params gather False |
|
[2023-12-11 05:39:51,206] [INFO] [logging.py:96:log_dist] [Rank 0] Creating torch.float16 ZeRO stage 3 optimizer |
|
[2023-12-11 05:39:51,330] [INFO] [utils.py:795:see_memory_usage] Stage 3 initialize beginning |
|
[2023-12-11 05:39:51,331] [INFO] [utils.py:796:see_memory_usage] MA 4.37 GB Max_MA 4.75 GB CA 8.35 GB Max_CA 8 GB |
|
[2023-12-11 05:39:51,331] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 98.11 GB, percent = 39.0% |
|
[2023-12-11 05:39:51,333] [INFO] [stage3.py:127:__init__] Reduce bucket size 500,000,000 |
|
[2023-12-11 05:39:51,333] [INFO] [stage3.py:128:__init__] Prefetch bucket size 30000000 |
|
[2023-12-11 05:39:51,450] [INFO] [utils.py:795:see_memory_usage] DeepSpeedZeRoOffload initialize [begin] |
|
[2023-12-11 05:39:51,450] [INFO] [utils.py:796:see_memory_usage] MA 4.37 GB Max_MA 4.37 GB CA 8.35 GB Max_CA 8 GB |
|
[2023-12-11 05:39:51,450] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 98.13 GB, percent = 39.0% |
|
Parameter Offload: Total persistent parameters: 266240 in 65 params |
|
[2023-12-11 05:39:51,757] [INFO] [utils.py:795:see_memory_usage] DeepSpeedZeRoOffload initialize [end] |
|
[2023-12-11 05:39:51,758] [INFO] [utils.py:796:see_memory_usage] MA 3.54 GB Max_MA 4.43 GB CA 8.35 GB Max_CA 8 GB |
|
[2023-12-11 05:39:51,758] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 98.13 GB, percent = 39.0% |
|
[2023-12-11 05:39:51,866] [INFO] [utils.py:795:see_memory_usage] Before creating fp16 partitions |
|
[2023-12-11 05:39:51,866] [INFO] [utils.py:796:see_memory_usage] MA 3.54 GB Max_MA 3.54 GB CA 8.35 GB Max_CA 8 GB |
|
[2023-12-11 05:39:51,866] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 98.13 GB, percent = 39.0% |
|
[2023-12-11 05:39:52,568] [INFO] [utils.py:795:see_memory_usage] After creating fp16 partitions: 3 |
|
[2023-12-11 05:39:52,569] [INFO] [utils.py:796:see_memory_usage] MA 3.54 GB Max_MA 3.54 GB CA 5.29 GB Max_CA 8 GB |
|
[2023-12-11 05:39:52,569] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 98.22 GB, percent = 39.0% |
|
[2023-12-11 05:39:52,708] [INFO] [utils.py:795:see_memory_usage] Before creating fp32 partitions |
|
[2023-12-11 05:39:52,709] [INFO] [utils.py:796:see_memory_usage] MA 3.54 GB Max_MA 3.54 GB CA 5.29 GB Max_CA 5 GB |
|
[2023-12-11 05:39:52,709] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 96.08 GB, percent = 38.2% |
|
[2023-12-11 05:39:52,831] [INFO] [utils.py:795:see_memory_usage] After creating fp32 partitions |
|
[2023-12-11 05:39:52,832] [INFO] [utils.py:796:see_memory_usage] MA 4.08 GB Max_MA 4.23 GB CA 5.99 GB Max_CA 6 GB |
|
[2023-12-11 05:39:52,832] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 96.11 GB, percent = 38.2% |
|
[2023-12-11 05:39:52,942] [INFO] [utils.py:795:see_memory_usage] Before initializing optimizer states |
|
[2023-12-11 05:39:52,942] [INFO] [utils.py:796:see_memory_usage] MA 4.08 GB Max_MA 4.08 GB CA 5.99 GB Max_CA 6 GB |
|
[2023-12-11 05:39:52,943] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 96.11 GB, percent = 38.2% |
|
[2023-12-11 05:39:53,083] [INFO] [utils.py:795:see_memory_usage] After initializing optimizer states |
|
[2023-12-11 05:39:53,084] [INFO] [utils.py:796:see_memory_usage] MA 5.17 GB Max_MA 5.47 GB CA 7.38 GB Max_CA 7 GB |
|
[2023-12-11 05:39:53,084] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 96.07 GB, percent = 38.2% |
|
[2023-12-11 05:39:53,084] [INFO] [stage3.py:479:_setup_for_real_optimizer] optimizer state initialized |
|
[2023-12-11 05:39:53,479] [INFO] [utils.py:795:see_memory_usage] After initializing ZeRO optimizer |
|
[2023-12-11 05:39:53,480] [INFO] [utils.py:796:see_memory_usage] MA 6.37 GB Max_MA 6.86 GB CA 9.05 GB Max_CA 9 GB |
|
[2023-12-11 05:39:53,480] [INFO] [utils.py:803:see_memory_usage] CPU Virtual Memory: used = 96.07 GB, percent = 38.2% |
|
[2023-12-11 05:39:53,480] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed Final Optimizer = FusedAdam |
|
[2023-12-11 05:39:53,480] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed using client LR scheduler |
|
[2023-12-11 05:39:53,480] [INFO] [logging.py:96:log_dist] [Rank 0] DeepSpeed LR Scheduler = <torch.optim.lr_scheduler.LambdaLR object at 0x7f53cfda67d0> |
|
[2023-12-11 05:39:53,480] [INFO] [logging.py:96:log_dist] [Rank 0] step=0, skipped=0, lr=[9.65e-06, 0.0005, 9.65e-06], mom=[(0.9, 0.95), (0.9, 0.95), (0.9, 0.95)] |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:979:print] DeepSpeedEngine configuration: |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] activation_checkpointing_config { |
|
"partition_activations": false, |
|
"contiguous_memory_optimization": false, |
|
"cpu_checkpointing": false, |
|
"number_checkpoints": null, |
|
"synchronize_checkpoint_boundary": false, |
|
"profile": false |
|
} |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'thread_count': 1, 'single_submit': False, 'overlap_events': True} |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] amp_enabled .................. False |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] amp_params ................... False |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983: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 |
|
} |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] bfloat16_enabled ............. False |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] checkpoint_parallel_write_pipeline False |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] checkpoint_tag_validation_enabled True |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] checkpoint_tag_validation_fail False |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] comms_config ................. <deepspeed.comm.config.DeepSpeedCommsConfig object at 0x7f53cf91c7d0> |
|
[2023-12-11 05:39:53,482] [INFO] [config.py:983:print] communication_data_type ...... None |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983: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}} |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] curriculum_enabled_legacy .... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] curriculum_params_legacy ..... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983: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}}}} |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] data_efficiency_enabled ...... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] dataloader_drop_last ......... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] disable_allgather ............ False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] dump_state ................... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] dynamic_loss_scale_args ...... {'init_scale': 65536, 'scale_window': 100, 'delayed_shift': 2, 'consecutive_hysteresis': False, 'min_scale': 1} |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_enabled ........... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_gas_boundary_resolution 1 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_layer_name ........ bert.encoder.layer |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_layer_num ......... 0 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_max_iter .......... 100 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_stability ......... 1e-06 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_tol ............... 0.01 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] eigenvalue_verbose ........... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] elasticity_enabled ........... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983: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 |
|
} |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] fp16_auto_cast ............... False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] fp16_enabled ................. True |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] fp16_master_weights_and_gradients False |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] global_rank .................. 0 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] grad_accum_dtype ............. None |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] gradient_accumulation_steps .. 1 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] gradient_clipping ............ 1.0 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] gradient_predivide_factor .... 1.0 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983: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 |
|
[2023-12-11 05:39:53,483] [INFO] [config.py:983:print] initial_dynamic_scale ........ 65536 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] load_universal_checkpoint .... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] loss_scale ................... 0 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] memory_breakdown ............. False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] mics_hierarchial_params_gather False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] mics_shard_size .............. -1 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] monitor_config ............... tensorboard=TensorBoardConfig(enabled=False, output_path='step1_tensorboard/ds_tensorboard_logs/', job_name='step1_model_tensorboard') wandb=WandbConfig(enabled=False, group=None, team=None, project='deepspeed') csv_monitor=CSVConfig(enabled=False, output_path='', job_name='DeepSpeedJobName') enabled=False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983: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 |
|
} |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] optimizer_legacy_fusion ...... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] optimizer_name ............... None |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] optimizer_params ............. None |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] pipeline ..................... {'stages': 'auto', 'partition': 'best', 'seed_layers': False, 'activation_checkpoint_interval': 0, 'pipe_partitioned': True, 'grad_partitioned': True} |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] pld_enabled .................. False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] pld_params ................... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] prescale_gradients ........... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] scheduler_name ............... None |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] scheduler_params ............. None |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] seq_parallel_communication_data_type torch.float32 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] sparse_attention ............. None |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] sparse_gradients_enabled ..... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] steps_per_print .............. 10 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] train_batch_size ............. 32 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] train_micro_batch_size_per_gpu 8 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] use_data_before_expert_parallel_ False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] use_node_local_storage ....... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] wall_clock_breakdown ......... False |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] weight_quantization_config ... None |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] world_size ................... 4 |
|
[2023-12-11 05:39:53,484] [INFO] [config.py:983:print] zero_allow_untested_optimizer False |
|
[2023-12-11 05:39:53,485] [INFO] [config.py:983:print] zero_config .................. stage=3 contiguous_gradients=True reduce_scatter=True reduce_bucket_size=500,000,000 use_multi_rank_bucket_allreduce=True allgather_partitions=True allgather_bucket_size=500,000,000 overlap_comm=True load_from_fp32_weights=True elastic_checkpoint=False offload_param=DeepSpeedZeroOffloadParamConfig(device='none', nvme_path=None, buffer_count=5, buffer_size=100,000,000, max_in_cpu=1,000,000,000, pin_memory=False) offload_optimizer=DeepSpeedZeroOffloadOptimizerConfig(device='none', nvme_path=None, buffer_count=4, pin_memory=False, 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=30000000 param_persistence_threshold=10000 model_persistence_threshold=sys.maxsize max_live_parameters=30000000 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=False pipeline_loading_checkpoint=False override_module_apply=True |
|
[2023-12-11 05:39:53,485] [INFO] [config.py:983:print] zero_enabled ................. True |
|
[2023-12-11 05:39:53,485] [INFO] [config.py:983:print] zero_force_ds_cpu_optimizer .. True |
|
[2023-12-11 05:39:53,485] [INFO] [config.py:983:print] zero_optimization_stage ...... 3 |
|
[2023-12-11 05:39:53,485] [INFO] [config.py:969:print_user_config] json = { |
|
"train_batch_size": 32, |
|
"train_micro_batch_size_per_gpu": 8, |
|
"steps_per_print": 10, |
|
"zero_optimization": { |
|
"stage": 3, |
|
"offload_param": { |
|
"device": "none" |
|
}, |
|
"offload_optimizer": { |
|
"device": "none" |
|
}, |
|
"stage3_param_persistence_threshold": 1.000000e+04, |
|
"stage3_max_live_parameters": 3.000000e+07, |
|
"stage3_prefetch_bucket_size": 3.000000e+07, |
|
"memory_efficient_linear": false |
|
}, |
|
"fp16": { |
|
"enabled": true, |
|
"loss_scale_window": 100 |
|
}, |
|
"gradient_clipping": 1.0, |
|
"prescale_gradients": false, |
|
"wall_clock_breakdown": false, |
|
"hybrid_engine": { |
|
"enabled": false, |
|
"max_out_tokens": 512, |
|
"inference_tp_size": 1, |
|
"release_inference_cache": false, |
|
"pin_parameters": true, |
|
"tp_gather_partition_size": 8 |
|
}, |
|
"tensorboard": { |
|
"enabled": false, |
|
"output_path": "step1_tensorboard/ds_tensorboard_logs/", |
|
"job_name": "step1_model_tensorboard" |
|
} |
|
} |
|
***** Running training ***** |
|
***** Evaluating perplexity, Epoch 0/3 ***** |
|
ppl: 4.460639476776123, loss: 1.4952921867370605 |
|
Beginning of Epoch 1/3, Total Micro Batches 13 |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/torch/utils/checkpoint.py:429: UserWarning: torch.utils.checkpoint: please pass in use_reentrant=True or use_reentrant=False explicitly. The default value of use_reentrant will be updated to be False in the future. To maintain current behavior, pass use_reentrant=True. It is recommended that you use use_reentrant=False. Refer to docs for more details on the differences between the two variants. |
|
warnings.warn( |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/torch/utils/checkpoint.py:429: UserWarning: torch.utils.checkpoint: please pass in use_reentrant=True or use_reentrant=False explicitly. The default value of use_reentrant will be updated to be False in the future. To maintain current behavior, pass use_reentrant=True. It is recommended that you use use_reentrant=False. Refer to docs for more details on the differences between the two variants. |
|
warnings.warn( |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/torch/utils/checkpoint.py:429: UserWarning: torch.utils.checkpoint: please pass in use_reentrant=True or use_reentrant=False explicitly. The default value of use_reentrant will be updated to be False in the future. To maintain current behavior, pass use_reentrant=True. It is recommended that you use use_reentrant=False. Refer to docs for more details on the differences between the two variants. |
|
warnings.warn( |
|
/home/t-sokumar/miniconda3/envs/ft/lib/python3.11/site-packages/torch/utils/checkpoint.py:429: UserWarning: torch.utils.checkpoint: please pass in use_reentrant=True or use_reentrant=False explicitly. The default value of use_reentrant will be updated to be False in the future. To maintain current behavior, pass use_reentrant=True. It is recommended that you use use_reentrant=False. Refer to docs for more details on the differences between the two variants. |
|
warnings.warn( |
|
Model Parameters: 6.927 B, Latency: 4.07s, TFLOPs: 10.30, Samples/sec: 1.97, Time/seq 0.51s, Batch Size: 8, Sequence Length: 512 |
|
Invalidate trace cache @ step 0: expected module 6, but got module 0 |
|
Model Parameters: 6.927 B, Latency: 3.74s, TFLOPs: 11.21, Samples/sec: 2.14, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.75s, TFLOPs: 11.17, Samples/sec: 2.13, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.73s, TFLOPs: 11.24, Samples/sec: 2.15, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.62s, TFLOPs: 11.55, Samples/sec: 2.21, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.62s, TFLOPs: 11.57, Samples/sec: 2.21, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.55, Samples/sec: 2.21, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.62s, TFLOPs: 11.56, Samples/sec: 2.21, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
[2023-12-11 05:40:33,349] [INFO] [logging.py:96:log_dist] [Rank 0] step=10, skipped=0, lr=[8.167395005683819e-06, 0.00042318108837739987, 8.167395005683819e-06], mom=[(0.9, 0.95), (0.9, 0.95), (0.9, 0.95)] |
|
[2023-12-11 05:40:33,349] [INFO] [timer.py:260:stop] epoch=0/micro_step=10/global_step=10, RunningAvgSamplesPerSec=8.762510054419295, CurrSamplesPerSec=8.82404309088149, MemAllocated=6.88GB, MaxMemAllocated=10.68GB |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.54, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.62s, TFLOPs: 11.55, Samples/sec: 2.21, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.62s, TFLOPs: 11.56, Samples/sec: 2.21, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.24s, TFLOPs: 12.92, Samples/sec: 2.47, Time/seq 0.40s, Batch Size: 8, Sequence Length: 512 |
|
***** Evaluating perplexity, Epoch 1/3 ***** |
|
Invalidate trace cache @ step 0: expected module 0, but got module 6 |
|
ppl: 1.7355413436889648, loss: 0.551319420337677 |
|
Beginning of Epoch 2/3, Total Micro Batches 13 |
|
Model Parameters: 6.927 B, Latency: 3.75s, TFLOPs: 11.17, Samples/sec: 2.13, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.76s, TFLOPs: 11.14, Samples/sec: 2.13, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.54, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
[2023-12-11 05:41:11,363] [INFO] [logging.py:96:log_dist] [Rank 0] step=20, skipped=0, lr=[4.6307168389720735e-06, 0.0002399335149726463, 4.6307168389720735e-06], mom=[(0.9, 0.95), (0.9, 0.95), (0.9, 0.95)] |
|
[2023-12-11 05:41:11,363] [INFO] [timer.py:260:stop] epoch=1/micro_step=7/global_step=20, RunningAvgSamplesPerSec=8.813860487355969, CurrSamplesPerSec=8.815178915300866, MemAllocated=6.88GB, MaxMemAllocated=11.06GB |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.24s, TFLOPs: 12.90, Samples/sec: 2.47, Time/seq 0.41s, Batch Size: 8, Sequence Length: 512 |
|
***** Evaluating perplexity, Epoch 2/3 ***** |
|
Invalidate trace cache @ step 0: expected module 0, but got module 6 |
|
ppl: 1.0645378828048706, loss: 0.0625406950712204 |
|
Beginning of Epoch 3/3, Total Micro Batches 13 |
|
Model Parameters: 6.927 B, Latency: 3.75s, TFLOPs: 11.16, Samples/sec: 2.13, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.77s, TFLOPs: 11.11, Samples/sec: 2.12, Time/seq 0.47s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
[2023-12-11 05:41:49,440] [INFO] [logging.py:96:log_dist] [Rank 0] step=30, skipped=0, lr=[1.2134356400744368e-06, 6.28723129572247e-05, 1.2134356400744368e-06], mom=[(0.9, 0.95), (0.9, 0.95), (0.9, 0.95)] |
|
[2023-12-11 05:41:49,441] [INFO] [timer.py:260:stop] epoch=2/micro_step=4/global_step=30, RunningAvgSamplesPerSec=8.8235289391519, CurrSamplesPerSec=8.771353693343526, MemAllocated=6.88GB, MaxMemAllocated=11.06GB |
|
Model Parameters: 6.927 B, Latency: 3.65s, TFLOPs: 11.47, Samples/sec: 2.19, Time/seq 0.46s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.53, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.64s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.64s, TFLOPs: 11.51, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.65s, TFLOPs: 11.48, Samples/sec: 2.19, Time/seq 0.46s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.64s, TFLOPs: 11.50, Samples/sec: 2.20, Time/seq 0.46s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.63s, TFLOPs: 11.52, Samples/sec: 2.20, Time/seq 0.45s, Batch Size: 8, Sequence Length: 512 |
|
Model Parameters: 6.927 B, Latency: 3.26s, TFLOPs: 12.83, Samples/sec: 2.45, Time/seq 0.41s, Batch Size: 8, Sequence Length: 512 |
|
***** Evaluating perplexity, Epoch 3/3 ***** |
|
Invalidate trace cache @ step 0: expected module 0, but got module 6 |
|
ppl: 1.031400442123413, loss: 0.03091755136847496 |
|
saving the final model ... |
|
[2023-12-11 05:42:35,188] [INFO] [launch.py:347:main] Process 1247715 exits successfully. |
|
[2023-12-11 05:42:35,188] [INFO] [launch.py:347:main] Process 1247716 exits successfully. |
|
[2023-12-11 05:42:35,189] [INFO] [launch.py:347:main] Process 1247717 exits successfully. |
|
[2023-12-11 05:44:14,200] [INFO] [launch.py:347:main] Process 1247714 exits successfully. |
|
|
