File size: 82,992 Bytes
2f85276 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 |
nohup: ignoring input
[2023-02-20 17:05:49,355] [WARNING] [runner.py:186:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only.
[2023-02-20 17:05:49,405] [INFO] [runner.py:548:main] cmd = /opt/conda/bin/python3 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNl19 --master_addr=127.0.0.1 --master_port=29500 --enable_each_rank_log=None tune_gpt.py --deepspeed deepspeed.json --upload-experiment
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
[2023-02-20 17:05:51,897] [INFO] [launch.py:135:main] 0 NCCL_VERSION=2.11.4
[2023-02-20 17:05:51,897] [INFO] [launch.py:142:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6]}
[2023-02-20 17:05:51,897] [INFO] [launch.py:148:main] nnodes=1, num_local_procs=7, node_rank=0
[2023-02-20 17:05:51,897] [INFO] [launch.py:161:main] global_rank_mapping=defaultdict(<class 'list'>, {'localhost': [0, 1, 2, 3, 4, 5, 6]})
[2023-02-20 17:05:51,897] [INFO] [launch.py:162:main] dist_world_size=7
[2023-02-20 17:05:51,897] [INFO] [launch.py:164:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
No config specified, defaulting to: apps/all
No config specified, defaulting to: apps/all
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
No config specified, defaulting to: apps/all
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
No config specified, defaulting to: apps/all
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
No config specified, defaulting to: apps/all
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
No config specified, defaulting to: apps/all
No config specified, defaulting to: apps/all
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
Found cached dataset apps (/home/.cache/huggingface/datasets/codeparrot___apps/all/0.0.0/04ac807715d07d6e5cc580f59cdc8213cd7dc4529d0bb819cca72c9f8e8c1aa5)
Max length: 2048
PyTorch: setting up devices
Max length: 2048
PyTorch: setting up devices
Max length: 2048
PyTorch: setting up devices
Max length: 2048
PyTorch: setting up devices
[2023-02-20 17:06:11,414] [INFO] [comm.py:657:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl
Max length: 2048
PyTorch: setting up devices
Max length: 2048
PyTorch: setting up devices
Max length: 2048
PyTorch: setting up devices
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
GPU memory occupied: 6883 MB.
GPU memory occupied: 6883 MB.
GPU memory occupied: 6883 MB.
GPU memory occupied: 6883 MB.
GPU memory occupied: 6883 MB.
GPU memory occupied: 6883 MB.
GPU memory occupied: 6883 MB.
[2023-02-20 17:06:12,424] [INFO] [logging.py:75:log_dist] [Rank 0] DeepSpeed info: version=0.8.1, git-hash=unknown, git-branch=unknown
[2023-02-20 17:06:14,006] [INFO] [logging.py:75:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Installed CUDA version 11.6 does not match the version torch was compiled with 11.7 but since the APIs are compatible, accepting this combination
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Detected CUDA files, patching ldflags
Emitting ninja build file /home/.cache/torch_extensions/py38_cu117/cpu_adam/build.ninja...
Building extension module cpu_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.825831890106201 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.6894984245300293 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.815955877304077 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.816244125366211 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.7123100757598877 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.8215184211730957 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.789081573486328 seconds
Adam Optimizer #0 is created with AVX2 arithmetic capability.
Config: alpha=0.000050, betas=(0.900000, 0.999000), weight_decay=0.050000, adam_w=1
[2023-02-20 17:06:19,789] [INFO] [logging.py:75:log_dist] [Rank 0] Using DeepSpeed Optimizer param name adamw as basic optimizer
[2023-02-20 17:06:19,794] [INFO] [logging.py:75:log_dist] [Rank 0] DeepSpeed Basic Optimizer = DeepSpeedCPUAdam
[2023-02-20 17:06:19,795] [INFO] [utils.py:53:is_zero_supported_optimizer] Checking ZeRO support for optimizer=DeepSpeedCPUAdam type=<class 'deepspeed.ops.adam.cpu_adam.DeepSpeedCPUAdam'>
[2023-02-20 17:06:19,795] [INFO] [logging.py:75:log_dist] [Rank 0] Creating torch.float32 ZeRO stage 2 optimizer
[2023-02-20 17:06:19,795] [INFO] [stage_1_and_2.py:144:__init__] Reduce bucket size 500000000
[2023-02-20 17:06:19,795] [INFO] [stage_1_and_2.py:145:__init__] Allgather bucket size 500000000
[2023-02-20 17:06:19,795] [INFO] [stage_1_and_2.py:146:__init__] CPU Offload: True
[2023-02-20 17:06:19,795] [INFO] [stage_1_and_2.py:147:__init__] Round robin gradient partitioning: False
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Emitting ninja build file /home/.cache/torch_extensions/py38_cu117/utils/build.ninja...
Building extension module utils...
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 utils...
Time to load utils op: 0.34289026260375977 seconds
Loading extension module utils...
Time to load utils op: 0.2027883529663086 seconds
Loading extension module utils...
Loading extension module utils...
Time to load utils op: 0.2021796703338623 seconds
Loading extension module utils...
Time to load utils op: 0.2025763988494873 seconds
Loading extension module utils...
Time to load utils op: 0.2033846378326416 seconds
Time to load utils op: 0.2029557228088379 seconds
Loading extension module utils...
Time to load utils op: 0.30292582511901855 seconds
Rank: 6 partition count [7] and sizes[(17885514, False)]
Rank: 5 partition count [7] and sizes[(17885514, False)]
Rank: 4 partition count [7] and sizes[(17885514, False)]
Rank: 2 partition count [7] and sizes[(17885514, False)]
Rank: 3 partition count [7] and sizes[(17885514, False)]
Rank: 1 partition count [7] and sizes[(17885514, False)]
Rank: 0 partition count [7] and sizes[(17885514, False)]
[2023-02-20 17:06:27,470] [INFO] [utils.py:825:see_memory_usage] Before initializing optimizer states
[2023-02-20 17:06:27,471] [INFO] [utils.py:826:see_memory_usage] MA 0.66 GB Max_MA 0.66 GB CA 0.85 GB Max_CA 1 GB
[2023-02-20 17:06:27,471] [INFO] [utils.py:834:see_memory_usage] CPU Virtual Memory: used = 39.85 GB, percent = 7.9%
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
Time to load utils op: 0.00165557861328125 seconds
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
Time to load utils op: 0.008014678955078125 seconds
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
Time to load utils op: 0.03653693199157715 seconds
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
Time to load utils op: 0.008858203887939453 seconds
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
Time to load utils op: 0.0007452964782714844 seconds
Time to load utils op: 0.046510934829711914 seconds
You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
[2023-02-20 17:06:28,120] [INFO] [utils.py:825:see_memory_usage] After initializing optimizer states
[2023-02-20 17:06:28,121] [INFO] [utils.py:826:see_memory_usage] MA 0.66 GB Max_MA 0.66 GB CA 0.85 GB Max_CA 1 GB
[2023-02-20 17:06:28,121] [INFO] [utils.py:834:see_memory_usage] CPU Virtual Memory: used = 40.4 GB, percent = 8.0%
[2023-02-20 17:06:28,121] [INFO] [stage_1_and_2.py:527:__init__] optimizer state initialized
[2023-02-20 17:06:28,222] [INFO] [utils.py:825:see_memory_usage] After initializing ZeRO optimizer
[2023-02-20 17:06:28,222] [INFO] [utils.py:826:see_memory_usage] MA 0.66 GB Max_MA 0.66 GB CA 0.85 GB Max_CA 1 GB
[2023-02-20 17:06:28,223] [INFO] [utils.py:834:see_memory_usage] CPU Virtual Memory: used = 40.4 GB, percent = 8.0%
[2023-02-20 17:06:28,223] [INFO] [logging.py:75:log_dist] [Rank 0] DeepSpeed Final Optimizer = adamw
[2023-02-20 17:06:28,223] [INFO] [logging.py:75:log_dist] [Rank 0] DeepSpeed using configured LR scheduler = WarmupLR
[2023-02-20 17:06:28,223] [INFO] [logging.py:75:log_dist] [Rank 0] DeepSpeed LR Scheduler = <deepspeed.runtime.lr_schedules.WarmupLR object at 0x7efce1a33f40>
[2023-02-20 17:06:28,224] [INFO] [logging.py:75:log_dist] [Rank 0] step=0, skipped=0, lr=[5e-05], mom=[[0.9, 0.999]]
[2023-02-20 17:06:28,226] [INFO] [config.py:1009:print] DeepSpeedEngine configuration:
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] activation_checkpointing_config {
"partition_activations": false,
"contiguous_memory_optimization": false,
"cpu_checkpointing": false,
"number_checkpoints": null,
"synchronize_checkpoint_boundary": false,
"profile": false
}
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'thread_count': 1, 'single_submit': False, 'overlap_events': True}
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] amp_enabled .................. False
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] amp_params ................... False
[2023-02-20 17:06:28,226] [INFO] [config.py:1013: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
}
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] bfloat16_enabled ............. False
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] checkpoint_parallel_write_pipeline False
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] checkpoint_tag_validation_enabled True
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] checkpoint_tag_validation_fail False
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] comms_config ................. <deepspeed.comm.config.DeepSpeedCommsConfig object at 0x7efd0eb74250>
[2023-02-20 17:06:28,226] [INFO] [config.py:1013:print] communication_data_type ...... None
[2023-02-20 17:06:28,227] [INFO] [config.py:1013: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}}
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] curriculum_enabled_legacy .... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] curriculum_params_legacy ..... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013: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}}}}
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] data_efficiency_enabled ...... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] dataloader_drop_last ......... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] disable_allgather ............ False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] dump_state ................... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] dynamic_loss_scale_args ...... None
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_enabled ........... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_gas_boundary_resolution 1
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_layer_name ........ bert.encoder.layer
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_layer_num ......... 0
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_max_iter .......... 100
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_stability ......... 1e-06
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_tol ............... 0.01
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] eigenvalue_verbose ........... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] elasticity_enabled ........... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] flops_profiler_config ........ {
"enabled": false,
"profile_step": 1,
"module_depth": -1,
"top_modules": 1,
"detailed": true,
"output_file": null
}
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] fp16_auto_cast ............... None
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] fp16_enabled ................. False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] fp16_master_weights_and_gradients False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] global_rank .................. 0
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] grad_accum_dtype ............. None
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] gradient_accumulation_steps .. 64
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] gradient_clipping ............ 1.0
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] gradient_predivide_factor .... 1.0
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] initial_dynamic_scale ........ 65536
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] load_universal_checkpoint .... False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] loss_scale ................... 0
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] memory_breakdown ............. False
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] monitor_config ............... tensorboard=TensorBoardConfig(enabled=True, output_path='logs/', job_name='train_neo') wandb=WandbConfig(enabled=False, group=None, team=None, project='deepspeed') csv_monitor=CSVConfig(enabled=False, output_path='', job_name='DeepSpeedJobName') enabled=True
[2023-02-20 17:06:28,227] [INFO] [config.py:1013: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
}
[2023-02-20 17:06:28,227] [INFO] [config.py:1013:print] optimizer_legacy_fusion ...... False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] optimizer_name ............... adamw
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] optimizer_params ............. {'lr': 5e-05, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.05}
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] pipeline ..................... {'stages': 'auto', 'partition': 'best', 'seed_layers': False, 'activation_checkpoint_interval': 0}
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] pld_enabled .................. False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] pld_params ................... False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] prescale_gradients ........... False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] scheduler_name ............... WarmupLR
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] scheduler_params ............. {'warmup_min_lr': 0, 'warmup_max_lr': 5e-05, 'warmup_num_steps': 500}
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] sparse_attention ............. None
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] sparse_gradients_enabled ..... False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] steps_per_print .............. 2000
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] train_batch_size ............. 2688
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] train_micro_batch_size_per_gpu 6
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] use_node_local_storage ....... False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] wall_clock_breakdown ......... False
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] world_size ................... 7
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] zero_allow_untested_optimizer True
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] zero_config .................. stage=2 contiguous_gradients=True reduce_scatter=True reduce_bucket_size=500000000 allgather_partitions=True allgather_bucket_size=500000000 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) 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
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] zero_enabled ................. True
[2023-02-20 17:06:28,228] [INFO] [config.py:1013:print] zero_optimization_stage ...... 2
[2023-02-20 17:06:28,228] [INFO] [config.py:998:print_user_config] json = {
"optimizer": {
"type": "AdamW",
"params": {
"lr": 5e-05,
"betas": [0.9, 0.999],
"eps": 1e-08,
"weight_decay": 0.05
}
},
"scheduler": {
"type": "WarmupLR",
"params": {
"warmup_min_lr": 0,
"warmup_max_lr": 5e-05,
"warmup_num_steps": 500
}
},
"zero_optimization": {
"stage": 2,
"offload_optimizer": {
"device": "cpu",
"pin_memory": true
},
"allgather_partitions": true,
"allgather_bucket_size": 5.000000e+08,
"overlap_comm": true,
"reduce_scatter": true,
"reduce_bucket_size": 5.000000e+08,
"contiguous_gradients": true
},
"tensorboard": {
"enabled": true,
"output_path": "logs/",
"job_name": "train_neo"
},
"zero_allow_untested_optimizer": true,
"gradient_accumulation_steps": 64,
"gradient_clipping": 1.0,
"steps_per_print": 2.000000e+03,
"train_batch_size": 2.688000e+03,
"train_micro_batch_size_per_gpu": 6,
"wall_clock_breakdown": false
}
Using /home/.cache/torch_extensions/py38_cu117 as PyTorch extensions root...
No modifications detected for re-loaded extension module utils, skipping build step...
Loading extension module utils...
Time to load utils op: 0.00042748451232910156 seconds
***** Running training *****
Num examples = 117232
Num Epochs = 10
Instantaneous batch size per device = 6
Total train batch size (w. parallel, distributed & accumulation) = 2688
Gradient Accumulation steps = 64
Total optimization steps = 430
Number of trainable parameters = 125198592
0%| | 0/430 [00:00<?, ?it/s]You're using a GPT2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
0%| | 1/430 [01:21<9:41:43, 81.36s/it]
{'loss': 6.8879, 'learning_rate': 0.0, 'epoch': 0.02}
0%| | 1/430 [01:21<9:41:43, 81.36s/it]
0%| | 2/430 [02:41<9:36:10, 80.77s/it]
1%| | 3/430 [04:02<9:33:33, 80.59s/it]
1%| | 4/430 [05:22<9:31:45, 80.53s/it]
1%| | 5/430 [06:42<9:30:07, 80.49s/it]
{'loss': 4.4038, 'learning_rate': 1.294882868674145e-05, 'epoch': 0.11}
1%| | 5/430 [06:42<9:30:07, 80.49s/it]
1%|▏ | 6/430 [08:03<9:28:34, 80.46s/it]
2%|▏ | 7/430 [09:24<9:28:07, 80.58s/it]
2%|▏ | 8/430 [10:44<9:26:23, 80.53s/it]
2%|▏ | 9/430 [12:05<9:24:58, 80.52s/it]
2%|▏ | 10/430 [13:25<9:23:23, 80.49s/it]
{'loss': 1.7286, 'learning_rate': 1.852558565662928e-05, 'epoch': 0.23}
2%|▏ | 10/430 [13:25<9:23:23, 80.49s/it]
3%|▎ | 11/430 [14:45<9:22:02, 80.48s/it]
3%|▎ | 12/430 [16:06<9:20:48, 80.50s/it]
3%|▎ | 13/430 [17:26<9:19:19, 80.48s/it]
3%|▎ | 14/430 [18:47<9:18:17, 80.52s/it]
3%|▎ | 15/430 [20:08<9:16:48, 80.50s/it]
{'loss': 0.5715, 'learning_rate': 2.1787779359648994e-05, 'epoch': 0.34}
3%|▎ | 15/430 [20:08<9:16:48, 80.50s/it]
4%|▎ | 16/430 [21:28<9:15:34, 80.52s/it]
4%|▍ | 17/430 [22:49<9:14:08, 80.51s/it]
4%|▍ | 18/430 [24:09<9:12:33, 80.47s/it]
4%|▍ | 19/430 [25:30<9:11:22, 80.49s/it]
5%|▍ | 20/430 [26:50<9:09:51, 80.47s/it]
{'loss': 0.5316, 'learning_rate': 2.41023426265171e-05, 'epoch': 0.46}
5%|▍ | 20/430 [26:50<9:09:51, 80.47s/it]
5%|▍ | 21/430 [28:10<9:08:25, 80.45s/it]
5%|▌ | 22/430 [29:31<9:08:02, 80.59s/it]
5%|▌ | 23/430 [30:52<9:06:16, 80.53s/it]
6%|▌ | 24/430 [32:12<9:04:45, 80.51s/it]
6%|▌ | 25/430 [33:32<9:03:06, 80.46s/it]
{'loss': 0.5132, 'learning_rate': 2.58976573734829e-05, 'epoch': 0.57}
6%|▌ | 25/430 [33:32<9:03:06, 80.46s/it]
6%|▌ | 26/430 [34:53<9:01:27, 80.42s/it]
6%|▋ | 27/430 [36:13<8:59:58, 80.39s/it]
7%|▋ | 28/430 [37:33<8:58:26, 80.36s/it]
7%|▋ | 29/430 [38:54<8:57:43, 80.46s/it]
7%|▋ | 30/430 [40:14<8:56:19, 80.45s/it]
{'loss': 0.4914, 'learning_rate': 2.7364536329536817e-05, 'epoch': 0.69}
7%|▋ | 30/430 [40:14<8:56:19, 80.45s/it]
7%|▋ | 31/430 [41:35<8:55:05, 80.46s/it]
7%|▋ | 32/430 [42:55<8:53:43, 80.46s/it]
8%|▊ | 33/430 [44:16<8:52:17, 80.45s/it]
8%|▊ | 34/430 [45:36<8:50:56, 80.45s/it]
8%|▊ | 35/430 [46:57<8:49:37, 80.45s/it]
{'loss': 0.4777, 'learning_rate': 2.8604764815275082e-05, 'epoch': 0.8}
8%|▊ | 35/430 [46:57<8:49:37, 80.45s/it]
8%|▊ | 36/430 [48:17<8:48:22, 80.46s/it]
9%|▊ | 37/430 [49:38<8:46:59, 80.46s/it]
9%|▉ | 38/430 [50:58<8:45:38, 80.46s/it]
9%|▉ | 39/430 [52:19<8:44:24, 80.47s/it]
9%|▉ | 40/430 [53:39<8:43:03, 80.47s/it]
{'loss': 0.4716, 'learning_rate': 2.9679099596404923e-05, 'epoch': 0.92}
9%|▉ | 40/430 [53:39<8:43:03, 80.47s/it]
10%|▉ | 41/430 [55:00<8:42:48, 80.64s/it]
10%|▉ | 42/430 [56:21<8:41:07, 80.59s/it]
10%|█ | 43/430 [57:41<8:39:31, 80.55s/it]
10%|█ | 44/430 [59:50<10:12:21, 95.19s/it]
10%|█ | 45/430 [1:01:11<9:42:25, 90.77s/it]
{'loss': 0.515, 'learning_rate': 3.0626730032556536e-05, 'epoch': 1.05}
10%|█ | 45/430 [1:01:11<9:42:25, 90.77s/it]
11%|█ | 46/430 [1:02:31<9:21:05, 87.67s/it]
11%|█ | 47/430 [1:03:52<9:06:42, 85.65s/it]
11%|█ | 48/430 [1:05:13<8:55:45, 84.15s/it]
11%|█▏ | 49/430 [1:06:33<8:47:15, 83.03s/it]
12%|█▏ | 50/430 [1:07:54<8:41:00, 82.27s/it]
{'loss': 0.4496, 'learning_rate': 3.147441434337073e-05, 'epoch': 1.16}
12%|█▏ | 50/430 [1:07:54<8:41:00, 82.27s/it]
12%|█▏ | 51/430 [1:09:14<8:36:14, 81.73s/it]
12%|█▏ | 52/430 [1:10:35<8:32:33, 81.36s/it]
12%|█▏ | 53/430 [1:11:55<8:30:00, 81.17s/it]
13%|█▎ | 54/430 [1:13:16<8:27:22, 80.96s/it]
13%|█▎ | 55/430 [1:14:37<8:25:21, 80.86s/it]
{'loss': 0.4428, 'learning_rate': 3.224123807782732e-05, 'epoch': 1.28}
13%|█▎ | 55/430 [1:14:37<8:25:21, 80.86s/it]
13%|█▎ | 56/430 [1:15:57<8:23:29, 80.77s/it]
13%|█▎ | 57/430 [1:17:18<8:21:34, 80.68s/it]
13%|█▎ | 58/430 [1:18:38<8:19:56, 80.63s/it]
14%|█▎ | 59/430 [1:19:59<8:18:24, 80.60s/it]
14%|█▍ | 60/430 [1:21:19<8:16:44, 80.55s/it]
{'loss': 0.4354, 'learning_rate': 3.294129329942464e-05, 'epoch': 1.39}
14%|█▍ | 60/430 [1:21:19<8:16:44, 80.55s/it]
14%|█▍ | 61/430 [1:22:40<8:15:19, 80.54s/it]
14%|█▍ | 62/430 [1:24:00<8:13:53, 80.53s/it]
15%|█▍ | 63/430 [1:25:21<8:12:48, 80.57s/it]
15%|█▍ | 64/430 [1:26:41<8:11:20, 80.55s/it]
15%|█▌ | 65/430 [1:28:02<8:09:50, 80.52s/it]
{'loss': 0.4267, 'learning_rate': 3.358528167653452e-05, 'epoch': 1.5}
15%|█▌ | 65/430 [1:28:02<8:09:50, 80.52s/it]
15%|█▌ | 66/430 [1:29:22<8:08:31, 80.52s/it]
16%|█▌ | 67/430 [1:30:43<8:07:03, 80.51s/it]
16%|█▌ | 68/430 [1:32:03<8:05:39, 80.50s/it]
16%|█▌ | 69/430 [1:33:24<8:04:17, 80.49s/it]
16%|█▋ | 70/430 [1:34:44<8:02:50, 80.47s/it]
{'loss': 0.4245, 'learning_rate': 3.4181521785162905e-05, 'epoch': 1.62}
16%|█▋ | 70/430 [1:34:44<8:02:50, 80.47s/it]
17%|█▋ | 71/430 [1:36:05<8:02:07, 80.58s/it]
17%|█▋ | 72/430 [1:37:25<8:00:27, 80.52s/it]
17%|█▋ | 73/430 [1:38:46<7:58:58, 80.50s/it]
17%|█▋ | 74/430 [1:40:06<7:57:28, 80.47s/it]
17%|█▋ | 75/430 [1:41:27<7:56:03, 80.46s/it]
{'loss': 0.4183, 'learning_rate': 3.473660804639045e-05, 'epoch': 1.73}
17%|█▋ | 75/430 [1:41:27<7:56:03, 80.46s/it]
18%|█▊ | 76/430 [1:42:47<7:54:59, 80.51s/it]
18%|█▊ | 77/430 [1:44:08<7:53:37, 80.50s/it]
18%|█▊ | 78/430 [1:45:28<7:52:21, 80.52s/it]
18%|█▊ | 79/430 [1:46:49<7:50:59, 80.51s/it]
19%|█▊ | 80/430 [1:48:09<7:49:36, 80.51s/it]
{'loss': 0.4186, 'learning_rate': 3.525585656629274e-05, 'epoch': 1.85}
19%|█▊ | 80/430 [1:48:09<7:49:36, 80.51s/it]
19%|█▉ | 81/430 [1:49:30<7:48:23, 80.53s/it]
19%|█▉ | 82/430 [1:50:50<7:47:05, 80.53s/it]
19%|█▉ | 83/430 [1:52:11<7:45:34, 80.50s/it]
20%|█▉ | 84/430 [1:53:32<7:45:20, 80.69s/it]
20%|█▉ | 85/430 [1:54:52<7:43:34, 80.62s/it]
{'loss': 0.4107, 'learning_rate': 3.574361557584177e-05, 'epoch': 1.96}
20%|█▉ | 85/430 [1:54:52<7:43:34, 80.62s/it]
20%|██ | 86/430 [1:56:13<7:42:01, 80.59s/it]
20%|██ | 87/430 [1:58:23<9:05:42, 95.46s/it]
20%|██ | 88/430 [1:59:44<8:38:35, 90.98s/it]
21%|██ | 89/430 [2:01:04<8:19:08, 87.82s/it]
21%|██ | 90/430 [2:02:25<8:05:18, 85.64s/it]
{'loss': 0.4564, 'learning_rate': 3.6292389118326696e-05, 'epoch': 2.09}
21%|██ | 90/430 [2:02:25<8:05:18, 85.64s/it]
21%|██ | 91/430 [2:03:45<7:55:07, 84.09s/it]
21%|██▏ | 92/430 [2:05:06<7:47:38, 83.01s/it]
22%|██▏ | 93/430 [2:06:26<7:42:05, 82.27s/it]
22%|██▏ | 94/430 [2:07:47<7:37:48, 81.75s/it]
22%|██▏ | 95/430 [2:09:07<7:34:41, 81.44s/it]
{'loss': 0.3989, 'learning_rate': 3.6722735522346654e-05, 'epoch': 2.21}
22%|██▏ | 95/430 [2:09:07<7:34:41, 81.44s/it]
22%|██▏ | 96/430 [2:10:28<7:31:43, 81.15s/it]
23%|██▎ | 97/430 [2:11:48<7:29:16, 80.95s/it]
23%|██▎ | 98/430 [2:13:09<7:27:04, 80.80s/it]
23%|██▎ | 99/430 [2:14:29<7:25:22, 80.73s/it]
23%|██▎ | 100/430 [2:15:50<7:23:40, 80.67s/it]
{'loss': 0.3972, 'learning_rate': 3.713122729342321e-05, 'epoch': 2.32}
23%|██▎ | 100/430 [2:15:50<7:23:40, 80.67s/it]
23%|██▎ | 101/430 [2:17:10<7:22:04, 80.62s/it]
24%|██▎ | 102/430 [2:18:31<7:20:28, 80.58s/it]
24%|██▍ | 103/430 [2:19:51<7:18:57, 80.54s/it]
24%|██▍ | 104/430 [2:21:12<7:17:41, 80.56s/it]
24%|██▍ | 105/430 [2:22:32<7:16:17, 80.55s/it]
{'loss': 0.3949, 'learning_rate': 3.751997728783617e-05, 'epoch': 2.44}
24%|██▍ | 105/430 [2:22:32<7:16:17, 80.55s/it]
25%|██▍ | 106/430 [2:23:53<7:15:01, 80.56s/it]
25%|██▍ | 107/430 [2:25:14<7:13:51, 80.59s/it]
25%|██▌ | 108/430 [2:26:34<7:12:27, 80.58s/it]
25%|██▌ | 109/430 [2:27:55<7:11:11, 80.60s/it]
26%|██▌ | 110/430 [2:29:15<7:09:47, 80.59s/it]
{'loss': 0.3924, 'learning_rate': 3.789080603898437e-05, 'epoch': 2.55}
26%|██▌ | 110/430 [2:29:15<7:09:47, 80.59s/it]
26%|██▌ | 111/430 [2:30:36<7:08:21, 80.57s/it]
26%|██▌ | 112/430 [2:31:57<7:07:20, 80.63s/it]
26%|██▋ | 113/430 [2:33:17<7:05:49, 80.60s/it]
27%|██▋ | 114/430 [2:34:38<7:04:22, 80.58s/it]
27%|██▋ | 115/430 [2:35:58<7:02:59, 80.57s/it]
{'loss': 0.3919, 'learning_rate': 3.8245293313935915e-05, 'epoch': 2.66}
27%|██▋ | 115/430 [2:35:58<7:02:59, 80.57s/it]
27%|██▋ | 116/430 [2:37:20<7:02:34, 80.75s/it]
27%|██▋ | 117/430 [2:38:40<7:00:55, 80.69s/it]
27%|██▋ | 118/430 [2:40:01<6:59:23, 80.65s/it]
28%|██▊ | 119/430 [2:41:21<6:57:54, 80.63s/it]
28%|██▊ | 120/430 [2:42:42<6:56:26, 80.60s/it]
{'loss': 0.3858, 'learning_rate': 3.8584818782171724e-05, 'epoch': 2.78}
28%|██▊ | 120/430 [2:42:42<6:56:26, 80.60s/it]
28%|██▊ | 121/430 [2:44:02<6:55:00, 80.58s/it]
28%|██▊ | 122/430 [2:45:23<6:53:30, 80.55s/it]
29%|██▊ | 123/430 [2:46:43<6:52:17, 80.58s/it]
29%|██▉ | 124/430 [2:48:04<6:51:04, 80.60s/it]
29%|██▉ | 125/430 [2:49:25<6:49:38, 80.59s/it]
{'loss': 0.3816, 'learning_rate': 3.8910594444236536e-05, 'epoch': 2.89}
29%|██▉ | 125/430 [2:49:25<6:49:38, 80.59s/it]
29%|██▉ | 126/430 [2:50:45<6:48:24, 80.61s/it]
30%|██▉ | 127/430 [2:52:06<6:47:00, 80.60s/it]
30%|██▉ | 128/430 [2:53:27<6:46:32, 80.77s/it]
30%|███ | 129/430 [2:54:48<6:44:56, 80.72s/it]
30%|███ | 130/430 [2:56:57<7:56:43, 95.35s/it]
{'loss': 0.4213, 'learning_rate': 3.922369074599331e-05, 'epoch': 3.02}
30%|███ | 130/430 [2:56:57<7:56:43, 95.35s/it]
30%|███ | 131/430 [2:58:18<7:32:58, 90.90s/it]
31%|███ | 132/430 [2:59:38<7:15:53, 87.76s/it]
31%|███ | 133/430 [3:00:59<7:03:46, 85.61s/it]
31%|███ | 134/430 [3:02:19<6:54:52, 84.10s/it]
31%|███▏ | 135/430 [3:03:40<6:48:13, 83.03s/it]
{'loss': 0.3764, 'learning_rate': 3.9525057798763787e-05, 'epoch': 3.14}
31%|███▏ | 135/430 [3:03:40<6:48:13, 83.03s/it]
32%|███▏ | 136/430 [3:05:00<6:43:08, 82.27s/it]
32%|███▏ | 137/430 [3:06:21<6:39:15, 81.76s/it]
32%|███▏ | 138/430 [3:07:41<6:36:06, 81.39s/it]
32%|███▏ | 139/430 [3:09:02<6:33:50, 81.20s/it]
33%|███▎ | 140/430 [3:10:23<6:31:28, 81.00s/it]
{'loss': 0.3702, 'learning_rate': 3.981554276636201e-05, 'epoch': 3.25}
33%|███▎ | 140/430 [3:10:23<6:31:28, 81.00s/it]
33%|███▎ | 141/430 [3:11:43<6:29:27, 80.86s/it]
33%|███▎ | 142/430 [3:13:04<6:27:32, 80.74s/it]
33%|███▎ | 143/430 [3:14:24<6:26:05, 80.72s/it]
33%|███▎ | 144/430 [3:15:45<6:24:25, 80.65s/it]
34%|███▎ | 145/430 [3:17:05<6:22:43, 80.57s/it]
{'loss': 0.3689, 'learning_rate': 4.0095904221004775e-05, 'epoch': 3.37}
34%|███▎ | 145/430 [3:17:05<6:22:43, 80.57s/it]
34%|███▍ | 146/430 [3:18:26<6:21:11, 80.53s/it]
34%|███▍ | 147/430 [3:19:46<6:19:49, 80.53s/it]
34%|███▍ | 148/430 [3:21:07<6:18:31, 80.54s/it]
35%|███▍ | 149/430 [3:22:27<6:17:16, 80.56s/it]
35%|███▍ | 150/430 [3:23:48<6:15:53, 80.55s/it]
{'loss': 0.3674, 'learning_rate': 4.0366824080900185e-05, 'epoch': 3.48}
35%|███▍ | 150/430 [3:23:48<6:15:53, 80.55s/it]Saving model checkpoint to ./results/checkpoint-150
Configuration saved in ./results/checkpoint-150/config.json
Model weights saved in ./results/checkpoint-150/pytorch_model.bin
tokenizer config file saved in ./results/checkpoint-150/tokenizer_config.json
Special tokens file saved in ./results/checkpoint-150/special_tokens_map.json
[2023-02-20 20:30:17,356] [INFO] [logging.py:75:log_dist] [Rank 0] [Torch] Checkpoint global_step151 is begin to save!
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py:1428: UserWarning: Positional args are being deprecated, use kwargs instead. Refer to https://pytorch.org/docs/master/generated/torch.nn.Module.html#torch.nn.Module.state_dict for details.
warnings.warn(
[2023-02-20 20:30:17,359] [INFO] [logging.py:75:log_dist] [Rank 0] Saving model checkpoint: ./results/checkpoint-150/global_step151/mp_rank_00_model_states.pt
[2023-02-20 20:30:17,359] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving ./results/checkpoint-150/global_step151/mp_rank_00_model_states.pt...
[2023-02-20 20:30:18,018] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved ./results/checkpoint-150/global_step151/mp_rank_00_model_states.pt.
[2023-02-20 20:30:18,019] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving ./results/checkpoint-150/global_step151/zero_pp_rank_0_mp_rank_00_optim_states.pt...
[2023-02-20 20:30:18,223] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved ./results/checkpoint-150/global_step151/zero_pp_rank_0_mp_rank_00_optim_states.pt.
[2023-02-20 20:30:18,223] [INFO] [engine.py:3407:_save_zero_checkpoint] zero checkpoint saved ./results/checkpoint-150/global_step151/zero_pp_rank_0_mp_rank_00_optim_states.pt
[2023-02-20 20:30:18,224] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step151 is ready now!
Deleting older checkpoint [results/checkpoint-45] due to args.save_total_limit
35%|███▌ | 151/430 [3:25:10<6:16:59, 81.07s/it]
35%|███▌ | 152/430 [3:26:31<6:14:42, 80.87s/it]
36%|███▌ | 153/430 [3:27:51<6:12:50, 80.76s/it]
36%|███▌ | 154/430 [3:29:11<6:11:03, 80.67s/it]
36%|███▌ | 155/430 [3:30:32<6:09:25, 80.60s/it]
{'loss': 0.3634, 'learning_rate': 4.062891760247626e-05, 'epoch': 3.6}
36%|███▌ | 155/430 [3:30:32<6:09:25, 80.60s/it]
36%|███▋ | 156/430 [3:31:52<6:07:55, 80.57s/it]
37%|███▋ | 157/430 [3:33:13<6:07:08, 80.69s/it]
37%|███▋ | 158/430 [3:34:34<6:05:33, 80.64s/it]
37%|███▋ | 159/430 [3:35:54<6:04:05, 80.61s/it]
37%|███▋ | 160/430 [3:37:15<6:02:37, 80.58s/it]
{'loss': 0.3622, 'learning_rate': 4.0882741795693975e-05, 'epoch': 3.71}
37%|███▋ | 160/430 [3:37:15<6:02:37, 80.58s/it]
37%|███▋ | 161/430 [3:38:36<6:01:14, 80.58s/it]
38%|███▊ | 162/430 [3:39:56<6:00:01, 80.60s/it]
38%|███▊ | 163/430 [3:41:17<5:59:32, 80.80s/it]
38%|███▊ | 164/430 [3:42:38<5:58:27, 80.86s/it]
38%|███▊ | 165/430 [3:43:59<5:56:38, 80.75s/it]
{'loss': 0.3545, 'learning_rate': 4.1128802551961496e-05, 'epoch': 3.83}
38%|███▊ | 165/430 [3:43:59<5:56:38, 80.75s/it]
39%|███▊ | 166/430 [3:45:20<5:55:05, 80.70s/it]
39%|███▉ | 167/430 [3:46:40<5:53:32, 80.66s/it]
39%|███▉ | 168/430 [3:48:01<5:52:04, 80.63s/it]
39%|███▉ | 169/430 [3:49:21<5:50:54, 80.67s/it]
40%|███▉ | 170/430 [3:50:42<5:49:25, 80.64s/it]
{'loss': 0.355, 'learning_rate': 4.136756071398985e-05, 'epoch': 3.94}
40%|███▉ | 170/430 [3:50:42<5:49:25, 80.64s/it]
40%|███▉ | 171/430 [3:52:03<5:48:04, 80.64s/it]
40%|████ | 172/430 [3:53:23<5:46:33, 80.59s/it]
40%|████ | 173/430 [3:55:33<6:48:29, 95.37s/it]
40%|████ | 174/430 [3:56:53<6:27:52, 90.91s/it]
41%|████ | 175/430 [3:58:14<6:13:17, 87.83s/it]
{'loss': 0.3978, 'learning_rate': 4.164502129979834e-05, 'epoch': 4.07}
41%|████ | 175/430 [3:58:14<6:13:17, 87.83s/it]
41%|████ | 176/430 [3:59:35<6:02:43, 85.68s/it]
41%|████ | 177/430 [4:00:55<5:54:43, 84.12s/it]
41%|████▏ | 178/430 [4:02:16<5:48:45, 83.04s/it]
42%|████▏ | 179/430 [4:03:36<5:44:09, 82.27s/it]
42%|████▏ | 180/430 [4:04:57<5:40:41, 81.77s/it]
{'loss': 0.3492, 'learning_rate': 4.186914608821452e-05, 'epoch': 4.18}
42%|████▏ | 180/430 [4:04:57<5:40:41, 81.77s/it]
42%|████▏ | 181/430 [4:06:17<5:37:50, 81.41s/it]
42%|████▏ | 182/430 [4:07:38<5:35:22, 81.14s/it]
43%|████▎ | 183/430 [4:08:58<5:33:16, 80.96s/it]
43%|████▎ | 184/430 [4:10:19<5:31:25, 80.83s/it]
43%|████▎ | 185/430 [4:11:40<5:29:40, 80.74s/it]
{'loss': 0.3483, 'learning_rate': 4.208719628018618e-05, 'epoch': 4.3}
43%|████▎ | 185/430 [4:11:40<5:29:40, 80.74s/it]
43%|████▎ | 186/430 [4:13:00<5:28:03, 80.67s/it]
43%|████▎ | 187/430 [4:14:21<5:26:44, 80.68s/it]
44%|████▎ | 188/430 [4:15:41<5:25:24, 80.68s/it]
44%|████▍ | 189/430 [4:17:02<5:23:51, 80.63s/it]
44%|████▍ | 190/430 [4:18:22<5:22:20, 80.58s/it]
{'loss': 0.3443, 'learning_rate': 4.229949249223448e-05, 'epoch': 4.41}
44%|████▍ | 190/430 [4:18:22<5:22:20, 80.58s/it]
44%|████▍ | 191/430 [4:19:43<5:20:53, 80.56s/it]
45%|████▍ | 192/430 [4:21:03<5:19:33, 80.56s/it]
45%|████▍ | 193/430 [4:22:24<5:18:45, 80.70s/it]
45%|████▌ | 194/430 [4:23:45<5:17:14, 80.65s/it]
45%|████▌ | 195/430 [4:25:06<5:15:47, 80.63s/it]
{'loss': 0.3448, 'learning_rate': 4.250633060899951e-05, 'epoch': 4.53}
45%|████▌ | 195/430 [4:25:06<5:15:47, 80.63s/it]
46%|████▌ | 196/430 [4:26:26<5:14:14, 80.57s/it]
46%|████▌ | 197/430 [4:27:47<5:12:46, 80.54s/it]
46%|████▌ | 198/430 [4:29:07<5:11:27, 80.55s/it]
46%|████▋ | 199/430 [4:30:28<5:10:06, 80.55s/it]
47%|████▋ | 200/430 [4:31:48<5:08:45, 80.55s/it]
{'loss': 0.3371, 'learning_rate': 4.2707984263311035e-05, 'epoch': 4.64}
47%|████▋ | 200/430 [4:31:48<5:08:45, 80.55s/it]
47%|████▋ | 201/430 [4:33:09<5:07:25, 80.55s/it]
47%|████▋ | 202/430 [4:34:29<5:06:03, 80.54s/it]
47%|████▋ | 203/430 [4:35:50<5:04:47, 80.56s/it]
47%|████▋ | 204/430 [4:37:10<5:03:25, 80.55s/it]
48%|████▊ | 205/430 [4:38:31<5:02:05, 80.56s/it]
{'loss': 0.3362, 'learning_rate': 4.290470701297542e-05, 'epoch': 4.76}
48%|████▊ | 205/430 [4:38:31<5:02:05, 80.56s/it]
48%|████▊ | 206/430 [4:39:52<5:00:48, 80.57s/it]
48%|████▊ | 207/430 [4:41:12<4:59:23, 80.55s/it]
48%|████▊ | 208/430 [4:42:33<4:58:04, 80.56s/it]
49%|████▊ | 209/430 [4:43:53<4:56:40, 80.54s/it]
49%|████▉ | 210/430 [4:45:14<4:55:18, 80.54s/it]
{'loss': 0.3333, 'learning_rate': 4.3096734257723994e-05, 'epoch': 4.87}
49%|████▉ | 210/430 [4:45:14<4:55:18, 80.54s/it]
49%|████▉ | 211/430 [4:46:35<4:54:27, 80.67s/it]
49%|████▉ | 212/430 [4:47:56<4:53:37, 80.81s/it]
50%|████▉ | 213/430 [4:49:16<4:52:03, 80.75s/it]
50%|████▉ | 214/430 [4:50:37<4:50:43, 80.75s/it]
50%|█████ | 215/430 [4:51:58<4:49:04, 80.67s/it]
{'loss': 0.3304, 'learning_rate': 4.3284284932676175e-05, 'epoch': 4.99}
50%|█████ | 215/430 [4:51:58<4:49:04, 80.67s/it]
50%|█████ | 216/430 [4:54:08<5:40:27, 95.45s/it]
50%|█████ | 217/430 [4:55:28<5:22:58, 90.98s/it]
51%|█████ | 218/430 [4:56:49<5:10:22, 87.84s/it]
51%|█████ | 219/430 [4:58:09<5:01:10, 85.64s/it]
51%|█████ | 220/430 [4:59:30<4:54:31, 84.15s/it]
{'loss': 0.3689, 'learning_rate': 4.350372288827778e-05, 'epoch': 5.11}
51%|█████ | 220/430 [4:59:30<4:54:31, 84.15s/it]
51%|█████▏ | 221/430 [5:00:50<4:49:24, 83.09s/it]
52%|█████▏ | 222/430 [5:02:11<4:45:37, 82.39s/it]
52%|█████▏ | 223/430 [5:03:32<4:42:20, 81.84s/it]
52%|█████▏ | 224/430 [5:04:52<4:39:35, 81.44s/it]
52%|█████▏ | 225/430 [5:06:13<4:37:20, 81.17s/it]
{'loss': 0.3281, 'learning_rate': 4.3682123980857955e-05, 'epoch': 5.23}
52%|█████▏ | 225/430 [5:06:13<4:37:20, 81.17s/it]
53%|█████▎ | 226/430 [5:07:33<4:35:24, 81.00s/it]
53%|█████▎ | 227/430 [5:08:54<4:33:33, 80.85s/it]
53%|█████▎ | 228/430 [5:10:14<4:31:47, 80.73s/it]
53%|█████▎ | 229/430 [5:11:35<4:30:14, 80.67s/it]
53%|█████▎ | 230/430 [5:12:55<4:28:43, 80.62s/it]
{'loss': 0.321, 'learning_rate': 4.3856654894696003e-05, 'epoch': 5.34}
53%|█████▎ | 230/430 [5:12:55<4:28:43, 80.62s/it]
54%|█████▎ | 231/430 [5:14:16<4:27:16, 80.58s/it]
54%|█████▍ | 232/430 [5:15:36<4:25:49, 80.55s/it]
54%|█████▍ | 233/430 [5:16:57<4:24:23, 80.53s/it]
54%|█████▍ | 234/430 [5:18:17<4:22:58, 80.50s/it]
55%|█████▍ | 235/430 [5:19:38<4:21:32, 80.48s/it]
{'loss': 0.3211, 'learning_rate': 4.402747998752177e-05, 'epoch': 5.46}
55%|█████▍ | 235/430 [5:19:38<4:21:32, 80.48s/it]
55%|█████▍ | 236/430 [5:20:58<4:20:16, 80.50s/it]
55%|█████▌ | 237/430 [5:22:19<4:18:53, 80.49s/it]
55%|█████▌ | 238/430 [5:23:39<4:17:33, 80.49s/it]
56%|█████▌ | 239/430 [5:25:00<4:16:10, 80.47s/it]
56%|█████▌ | 240/430 [5:26:20<4:14:47, 80.46s/it]
{'loss': 0.3177, 'learning_rate': 4.41947533645377e-05, 'epoch': 5.57}
56%|█████▌ | 240/430 [5:26:20<4:14:47, 80.46s/it]
56%|█████▌ | 241/430 [5:27:40<4:13:24, 80.45s/it]
56%|█████▋ | 242/430 [5:29:01<4:12:06, 80.46s/it]
57%|█████▋ | 243/430 [5:30:21<4:10:48, 80.47s/it]
57%|█████▋ | 244/430 [5:31:42<4:09:29, 80.48s/it]
57%|█████▋ | 245/430 [5:33:02<4:08:08, 80.48s/it]
{'loss': 0.3178, 'learning_rate': 4.435861971380601e-05, 'epoch': 5.69}
57%|█████▋ | 245/430 [5:33:02<4:08:08, 80.48s/it]
57%|█████▋ | 246/430 [5:34:23<4:06:46, 80.47s/it]
57%|█████▋ | 247/430 [5:35:44<4:05:59, 80.65s/it]
58%|█████▊ | 248/430 [5:37:04<4:04:28, 80.59s/it]
58%|█████▊ | 249/430 [5:38:25<4:03:06, 80.59s/it]
58%|█████▊ | 250/430 [5:39:46<4:01:45, 80.59s/it]
{'loss': 0.3147, 'learning_rate': 4.451921505824621e-05, 'epoch': 5.8}
58%|█████▊ | 250/430 [5:39:46<4:01:45, 80.59s/it]
58%|█████▊ | 251/430 [5:41:06<4:00:32, 80.63s/it]
59%|█████▊ | 252/430 [5:42:27<3:59:37, 80.77s/it]
59%|█████▉ | 253/430 [5:43:48<3:58:03, 80.70s/it]
59%|█████▉ | 254/430 [5:45:08<3:56:32, 80.64s/it]
59%|█████▉ | 255/430 [5:46:29<3:55:05, 80.60s/it]
{'loss': 0.3131, 'learning_rate': 4.467666743403693e-05, 'epoch': 5.92}
59%|█████▉ | 255/430 [5:46:29<3:55:05, 80.60s/it]
60%|█████▉ | 256/430 [5:47:50<3:54:05, 80.72s/it]
60%|█████▉ | 257/430 [5:49:10<3:52:33, 80.65s/it]
60%|██████ | 258/430 [5:50:31<3:51:09, 80.64s/it]
60%|██████ | 259/430 [5:52:40<4:31:29, 95.26s/it]
60%|██████ | 260/430 [5:54:01<4:17:26, 90.86s/it]
{'loss': 0.3454, 'learning_rate': 4.483109750389942e-05, 'epoch': 6.05}
60%|██████ | 260/430 [5:54:01<4:17:26, 90.86s/it]
61%|██████ | 261/430 [5:55:22<4:07:10, 87.75s/it]
61%|██████ | 262/430 [5:56:42<3:59:47, 85.64s/it]
61%|██████ | 263/430 [5:58:03<3:54:02, 84.09s/it]
61%|██████▏ | 264/430 [5:59:23<3:49:37, 83.00s/it]
62%|██████▏ | 265/430 [6:00:44<3:46:08, 82.23s/it]
{'loss': 0.3044, 'learning_rate': 4.498261911262221e-05, 'epoch': 6.16}
62%|██████▏ | 265/430 [6:00:44<3:46:08, 82.23s/it]
62%|██████▏ | 266/430 [6:02:04<3:43:19, 81.70s/it]
62%|██████▏ | 267/430 [6:03:25<3:40:59, 81.34s/it]
62%|██████▏ | 268/430 [6:04:45<3:39:08, 81.16s/it]
63%|██████▎ | 269/430 [6:06:06<3:37:16, 80.97s/it]
63%|██████▎ | 270/430 [6:07:26<3:35:33, 80.84s/it]
{'loss': 0.3066, 'learning_rate': 4.513133979123424e-05, 'epoch': 6.28}
63%|██████▎ | 270/430 [6:07:26<3:35:33, 80.84s/it]
63%|██████▎ | 271/430 [6:08:47<3:34:06, 80.80s/it]
63%|██████▎ | 272/430 [6:10:08<3:32:33, 80.72s/it]
63%|██████▎ | 273/430 [6:11:28<3:31:01, 80.65s/it]
64%|██████▎ | 274/430 [6:12:49<3:29:32, 80.60s/it]
64%|██████▍ | 275/430 [6:14:09<3:28:03, 80.54s/it]
{'loss': 0.301, 'learning_rate': 4.527736121541934e-05, 'epoch': 6.39}
64%|██████▍ | 275/430 [6:14:09<3:28:03, 80.54s/it]
64%|██████▍ | 276/430 [6:15:29<3:26:41, 80.53s/it]
64%|██████▍ | 277/430 [6:16:50<3:25:18, 80.51s/it]
65%|██████▍ | 278/430 [6:18:10<3:23:55, 80.50s/it]
65%|██████▍ | 279/430 [6:19:31<3:22:38, 80.52s/it]
65%|██████▌ | 280/430 [6:20:52<3:21:30, 80.60s/it]
{'loss': 0.2992, 'learning_rate': 4.5420779623067014e-05, 'epoch': 6.5}
65%|██████▌ | 280/430 [6:20:52<3:21:30, 80.60s/it]
65%|██████▌ | 281/430 [6:22:12<3:20:06, 80.58s/it]
66%|██████▌ | 282/430 [6:23:33<3:18:45, 80.58s/it]
66%|██████▌ | 283/430 [6:24:54<3:17:25, 80.58s/it]
66%|██████▌ | 284/430 [6:26:14<3:16:05, 80.59s/it]
66%|██████▋ | 285/430 [6:27:35<3:14:40, 80.56s/it]
{'loss': 0.2963, 'learning_rate': 4.55616861952542e-05, 'epoch': 6.62}
66%|██████▋ | 285/430 [6:27:35<3:14:40, 80.56s/it]
67%|██████▋ | 286/430 [6:28:55<3:13:19, 80.55s/it]
67%|██████▋ | 287/430 [6:30:16<3:11:55, 80.53s/it]
67%|██████▋ | 288/430 [6:31:37<3:10:52, 80.65s/it]
67%|██████▋ | 289/430 [6:32:57<3:09:26, 80.62s/it]
67%|██████▋ | 290/430 [6:34:18<3:07:59, 80.57s/it]
{'loss': 0.2912, 'learning_rate': 4.5700167404435284e-05, 'epoch': 6.73}
67%|██████▋ | 290/430 [6:34:18<3:07:59, 80.57s/it]
68%|██████▊ | 291/430 [6:35:39<3:06:56, 80.69s/it]
68%|██████▊ | 292/430 [6:36:59<3:05:26, 80.63s/it]
68%|██████▊ | 293/430 [6:38:20<3:04:03, 80.61s/it]
68%|██████▊ | 294/430 [6:39:40<3:02:37, 80.57s/it]
69%|██████▊ | 295/430 [6:41:01<3:01:16, 80.56s/it]
{'loss': 0.2916, 'learning_rate': 4.583630533316995e-05, 'epoch': 6.85}
69%|██████▊ | 295/430 [6:41:01<3:01:16, 80.56s/it]
69%|██████▉ | 296/430 [6:42:21<2:59:53, 80.55s/it]
69%|██████▉ | 297/430 [6:43:42<2:58:55, 80.72s/it]
69%|██████▉ | 298/430 [6:45:03<2:57:26, 80.66s/it]
70%|██████▉ | 299/430 [6:46:23<2:55:57, 80.59s/it]
70%|██████▉ | 300/430 [6:47:44<2:54:31, 80.55s/it]
{'loss': 0.2918, 'learning_rate': 4.597017796633075e-05, 'epoch': 6.96}
70%|██████▉ | 300/430 [6:47:44<2:54:31, 80.55s/it]Saving model checkpoint to ./results/checkpoint-300
Configuration saved in ./results/checkpoint-300/config.json
Model weights saved in ./results/checkpoint-300/pytorch_model.bin
tokenizer config file saved in ./results/checkpoint-300/tokenizer_config.json
Special tokens file saved in ./results/checkpoint-300/special_tokens_map.json
[2023-02-20 23:54:13,030] [INFO] [logging.py:75:log_dist] [Rank 0] [Torch] Checkpoint global_step303 is begin to save!
[2023-02-20 23:54:13,032] [INFO] [logging.py:75:log_dist] [Rank 0] Saving model checkpoint: ./results/checkpoint-300/global_step303/mp_rank_00_model_states.pt
[2023-02-20 23:54:13,032] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving ./results/checkpoint-300/global_step303/mp_rank_00_model_states.pt...
[2023-02-20 23:54:13,586] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved ./results/checkpoint-300/global_step303/mp_rank_00_model_states.pt.
[2023-02-20 23:54:13,587] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving ./results/checkpoint-300/global_step303/zero_pp_rank_0_mp_rank_00_optim_states.pt...
[2023-02-20 23:54:13,739] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved ./results/checkpoint-300/global_step303/zero_pp_rank_0_mp_rank_00_optim_states.pt.
[2023-02-20 23:54:13,740] [INFO] [engine.py:3407:_save_zero_checkpoint] zero checkpoint saved ./results/checkpoint-300/global_step303/zero_pp_rank_0_mp_rank_00_optim_states.pt
[2023-02-20 23:54:13,740] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step303 is ready now!
Deleting older checkpoint [results/checkpoint-150] due to args.save_total_limit
70%|███████ | 301/430 [6:49:06<2:54:08, 80.99s/it]
70%|███████ | 302/430 [6:51:15<3:24:02, 95.64s/it]
70%|███████ | 303/430 [6:52:36<3:12:48, 91.09s/it]
71%|███████ | 304/430 [6:53:56<3:04:36, 87.91s/it]
71%|███████ | 305/430 [6:55:17<2:58:30, 85.68s/it]
{'loss': 0.3224, 'learning_rate': 4.612793909203516e-05, 'epoch': 7.09}
71%|███████ | 305/430 [6:55:17<2:58:30, 85.68s/it]
71%|███████ | 306/430 [6:56:38<2:54:00, 84.19s/it]
71%|███████▏ | 307/430 [6:57:59<2:50:44, 83.29s/it]
72%|███████▏ | 308/430 [6:59:19<2:47:40, 82.46s/it]
72%|███████▏ | 309/430 [7:00:40<2:45:10, 81.91s/it]
72%|███████▏ | 310/430 [7:02:00<2:42:57, 81.48s/it]
{'loss': 0.2838, 'learning_rate': 4.6257084073957534e-05, 'epoch': 7.21}
72%|███████▏ | 310/430 [7:02:00<2:42:57, 81.48s/it]
72%|███████▏ | 311/430 [7:03:21<2:41:01, 81.18s/it]
73%|███████▎ | 312/430 [7:04:41<2:39:14, 80.97s/it]
73%|███████▎ | 313/430 [7:06:02<2:37:35, 80.82s/it]
73%|███████▎ | 314/430 [7:07:22<2:36:02, 80.71s/it]
73%|███████▎ | 315/430 [7:08:43<2:34:33, 80.64s/it]
{'loss': 0.2803, 'learning_rate': 4.6384188765246125e-05, 'epoch': 7.32}
73%|███████▎ | 315/430 [7:08:43<2:34:33, 80.64s/it]
73%|███████▎ | 316/430 [7:10:03<2:33:10, 80.62s/it]
74%|███████▎ | 317/430 [7:11:24<2:31:46, 80.59s/it]
74%|███████▍ | 318/430 [7:12:44<2:30:23, 80.56s/it]
74%|███████▍ | 319/430 [7:14:05<2:29:01, 80.56s/it]
74%|███████▍ | 320/430 [7:15:26<2:27:42, 80.57s/it]
{'loss': 0.2839, 'learning_rate': 4.6509316631405805e-05, 'epoch': 7.44}
74%|███████▍ | 320/430 [7:15:26<2:27:42, 80.57s/it]
75%|███████▍ | 321/430 [7:16:46<2:26:19, 80.54s/it]
75%|███████▍ | 322/430 [7:18:07<2:24:58, 80.54s/it]
75%|███████▌ | 323/430 [7:19:27<2:23:39, 80.55s/it]
75%|███████▌ | 324/430 [7:20:48<2:22:15, 80.53s/it]
76%|███████▌ | 325/430 [7:22:08<2:20:55, 80.53s/it]
{'loss': 0.2776, 'learning_rate': 4.663252822198809e-05, 'epoch': 7.55}
76%|███████▌ | 325/430 [7:22:08<2:20:55, 80.53s/it]
76%|███████▌ | 326/430 [7:23:29<2:19:34, 80.52s/it]
76%|███████▌ | 327/430 [7:24:49<2:18:11, 80.50s/it]
76%|███████▋ | 328/430 [7:26:10<2:16:52, 80.52s/it]
77%|███████▋ | 329/430 [7:27:30<2:15:32, 80.52s/it]
77%|███████▋ | 330/430 [7:28:51<2:14:13, 80.53s/it]
{'loss': 0.2763, 'learning_rate': 4.675388134653313e-05, 'epoch': 7.66}
77%|███████▋ | 330/430 [7:28:51<2:14:13, 80.53s/it]
77%|███████▋ | 331/430 [7:30:11<2:12:57, 80.58s/it]
77%|███████▋ | 332/430 [7:31:32<2:11:44, 80.66s/it]
77%|███████▋ | 333/430 [7:32:53<2:10:32, 80.75s/it]
78%|███████▊ | 334/430 [7:34:14<2:09:04, 80.67s/it]
78%|███████▊ | 335/430 [7:35:34<2:07:39, 80.62s/it]
{'loss': 0.2731, 'learning_rate': 4.687343123743873e-05, 'epoch': 7.78}
78%|███████▊ | 335/430 [7:35:34<2:07:39, 80.62s/it]
78%|███████▊ | 336/430 [7:36:55<2:06:30, 80.75s/it]
78%|███████▊ | 337/430 [7:38:16<2:05:08, 80.73s/it]
79%|███████▊ | 338/430 [7:39:36<2:03:41, 80.66s/it]
79%|███████▉ | 339/430 [7:40:57<2:02:15, 80.61s/it]
79%|███████▉ | 340/430 [7:42:17<2:00:50, 80.56s/it]
{'loss': 0.271, 'learning_rate': 4.699123070090503e-05, 'epoch': 7.89}
79%|███████▉ | 340/430 [7:42:17<2:00:50, 80.56s/it]
79%|███████▉ | 341/430 [7:43:38<1:59:27, 80.53s/it]
80%|███████▉ | 342/430 [7:44:58<1:58:06, 80.53s/it]
80%|███████▉ | 343/430 [7:46:19<1:56:47, 80.55s/it]
80%|████████ | 344/430 [7:47:40<1:55:30, 80.59s/it]
80%|████████ | 345/430 [7:49:49<2:14:57, 95.26s/it]
{'loss': 0.3009, 'learning_rate': 4.71073302569872e-05, 'epoch': 8.02}
80%|████████ | 345/430 [7:49:49<2:14:57, 95.26s/it]
80%|████████ | 346/430 [7:51:10<2:07:10, 90.84s/it]
81%|████████ | 347/430 [7:52:30<2:01:21, 87.73s/it]
81%|████████ | 348/430 [7:53:51<1:56:55, 85.56s/it]
81%|████████ | 349/430 [7:55:11<1:53:26, 84.03s/it]
81%|████████▏ | 350/430 [7:56:32<1:50:39, 83.00s/it]
{'loss': 0.265, 'learning_rate': 4.7221778269686165e-05, 'epoch': 8.14}
81%|████████▏ | 350/430 [7:56:32<1:50:39, 83.00s/it]
82%|████████▏ | 351/430 [7:57:52<1:48:17, 82.24s/it]
82%|████████▏ | 352/430 [7:59:13<1:46:14, 81.72s/it]
82%|████████▏ | 353/430 [8:00:34<1:44:40, 81.56s/it]
82%|████████▏ | 354/430 [8:01:54<1:42:53, 81.22s/it]
83%|████████▎ | 355/430 [8:03:15<1:41:14, 80.99s/it]
{'loss': 0.2651, 'learning_rate': 4.73346210679156e-05, 'epoch': 8.25}
83%|████████▎ | 355/430 [8:03:15<1:41:14, 80.99s/it]
83%|████████▎ | 356/430 [8:04:36<1:39:57, 81.05s/it]
83%|████████▎ | 357/430 [8:05:56<1:38:25, 80.90s/it]
83%|████████▎ | 358/430 [8:07:17<1:36:55, 80.77s/it]
83%|████████▎ | 359/430 [8:08:38<1:35:30, 80.71s/it]
84%|████████▎ | 360/430 [8:09:58<1:34:03, 80.63s/it]
{'loss': 0.2607, 'learning_rate': 4.744590305810234e-05, 'epoch': 8.37}
84%|████████▎ | 360/430 [8:09:58<1:34:03, 80.63s/it]
84%|████████▍ | 361/430 [8:11:18<1:32:39, 80.57s/it]
84%|████████▍ | 362/430 [8:12:39<1:31:17, 80.55s/it]
84%|████████▍ | 363/430 [8:13:59<1:29:56, 80.54s/it]
85%|████████▍ | 364/430 [8:15:20<1:28:34, 80.52s/it]
85%|████████▍ | 365/430 [8:16:40<1:27:12, 80.50s/it]
{'loss': 0.2599, 'learning_rate': 4.7555666829105464e-05, 'epoch': 8.48}
85%|████████▍ | 365/430 [8:16:40<1:27:12, 80.50s/it]
85%|████████▌ | 366/430 [8:18:01<1:25:49, 80.47s/it]
85%|████████▌ | 367/430 [8:19:21<1:24:27, 80.44s/it]
86%|████████▌ | 368/430 [8:20:42<1:23:10, 80.49s/it]
86%|████████▌ | 369/430 [8:22:02<1:21:48, 80.47s/it]
86%|████████▌ | 370/430 [8:23:23<1:20:29, 80.49s/it]
{'loss': 0.2564, 'learning_rate': 4.7663953250074004e-05, 'epoch': 8.6}
86%|████████▌ | 370/430 [8:23:23<1:20:29, 80.49s/it]
86%|████████▋ | 371/430 [8:24:43<1:19:09, 80.51s/it]
87%|████████▋ | 372/430 [8:26:04<1:17:55, 80.61s/it]
87%|████████▋ | 373/430 [8:27:25<1:16:32, 80.57s/it]
87%|████████▋ | 374/430 [8:28:45<1:15:10, 80.55s/it]
87%|████████▋ | 375/430 [8:30:06<1:13:49, 80.54s/it]
{'loss': 0.2584, 'learning_rate': 4.777080156180637e-05, 'epoch': 8.71}
87%|████████▋ | 375/430 [8:30:06<1:13:49, 80.54s/it]
87%|████████▋ | 376/430 [8:31:26<1:12:29, 80.54s/it]
88%|████████▊ | 377/430 [8:32:47<1:11:07, 80.53s/it]
88%|████████▊ | 378/430 [8:34:07<1:09:47, 80.53s/it]
88%|████████▊ | 379/430 [8:35:28<1:08:26, 80.52s/it]
88%|████████▊ | 380/430 [8:36:48<1:07:07, 80.55s/it]
{'loss': 0.2578, 'learning_rate': 4.7876249462122306e-05, 'epoch': 8.83}
88%|████████▊ | 380/430 [8:36:48<1:07:07, 80.55s/it]
89%|████████▊ | 381/430 [8:38:09<1:05:47, 80.56s/it]
89%|████████▉ | 382/430 [8:39:29<1:04:25, 80.53s/it]
89%|████████▉ | 383/430 [8:40:50<1:03:04, 80.53s/it]
89%|████████▉ | 384/430 [8:42:10<1:01:43, 80.51s/it]
90%|████████▉ | 385/430 [8:43:31<1:00:22, 80.49s/it]
{'loss': 0.2495, 'learning_rate': 4.798033318571224e-05, 'epoch': 8.94}
90%|████████▉ | 385/430 [8:43:31<1:00:22, 80.49s/it]
90%|████████▉ | 386/430 [8:44:52<59:06, 80.61s/it]
90%|█████████ | 387/430 [8:46:12<57:43, 80.55s/it]
90%|█████████ | 388/430 [8:48:22<1:06:42, 95.31s/it]
90%|█████████ | 389/430 [8:49:42<1:02:05, 90.85s/it]
91%|█████████ | 390/430 [8:51:03<58:29, 87.73s/it]
{'loss': 0.2808, 'learning_rate': 4.810348191078279e-05, 'epoch': 9.07}
91%|█████████ | 390/430 [8:51:03<58:29, 87.73s/it]
91%|█████████ | 391/430 [8:52:23<55:37, 85.58s/it]
91%|█████████ | 392/430 [8:53:44<53:14, 84.05s/it]
91%|█████████▏| 393/430 [8:55:04<51:10, 82.98s/it]
92%|█████████▏| 394/430 [8:56:25<49:25, 82.39s/it]
92%|█████████▏| 395/430 [8:57:46<47:43, 81.82s/it]
{'loss': 0.2463, 'learning_rate': 4.82046852530342e-05, 'epoch': 9.18}
92%|█████████▏| 395/430 [8:57:46<47:43, 81.82s/it]
92%|█████████▏| 396/430 [8:59:06<46:09, 81.45s/it]
92%|█████████▏| 397/430 [9:00:27<44:37, 81.15s/it]
93%|█████████▎| 398/430 [9:01:47<43:09, 80.93s/it]
93%|█████████▎| 399/430 [9:03:08<41:44, 80.80s/it]
93%|█████████▎| 400/430 [9:04:28<40:20, 80.70s/it]
{'loss': 0.2418, 'learning_rate': 4.830463137837162e-05, 'epoch': 9.3}
93%|█████████▎| 400/430 [9:04:28<40:20, 80.70s/it]
93%|█████████▎| 401/430 [9:05:49<38:58, 80.65s/it]
93%|█████████▎| 402/430 [9:07:09<37:38, 80.65s/it]
94%|█████████▎| 403/430 [9:08:30<36:16, 80.59s/it]
94%|█████████▍| 404/430 [9:09:50<34:54, 80.56s/it]
94%|█████████▍| 405/430 [9:11:11<33:33, 80.54s/it]
{'loss': 0.2451, 'learning_rate': 4.8403351139919656e-05, 'epoch': 9.41}
94%|█████████▍| 405/430 [9:11:11<33:33, 80.54s/it]
94%|█████████▍| 406/430 [9:12:32<32:14, 80.61s/it]
95%|█████████▍| 407/430 [9:13:52<30:52, 80.56s/it]
95%|█████████▍| 408/430 [9:15:12<29:31, 80.54s/it]
95%|█████████▌| 409/430 [9:16:33<28:11, 80.54s/it]
95%|█████████▌| 410/430 [9:17:53<26:50, 80.51s/it]
{'loss': 0.2403, 'learning_rate': 4.850087426881512e-05, 'epoch': 9.53}
95%|█████████▌| 410/430 [9:17:53<26:50, 80.51s/it]
96%|█████████▌| 411/430 [9:19:15<25:33, 80.70s/it]
96%|█████████▌| 412/430 [9:20:35<24:11, 80.65s/it]
96%|█████████▌| 413/430 [9:21:56<22:50, 80.61s/it]
96%|█████████▋| 414/430 [9:23:16<21:28, 80.56s/it]
97%|█████████▋| 415/430 [9:24:37<20:09, 80.62s/it]
{'loss': 0.2411, 'learning_rate': 4.859722942795827e-05, 'epoch': 9.64}
97%|█████████▋| 415/430 [9:24:37<20:09, 80.62s/it]
97%|█████████▋| 416/430 [9:25:57<18:48, 80.58s/it]
97%|█████████▋| 417/430 [9:27:18<17:27, 80.54s/it]
97%|█████████▋| 418/430 [9:28:38<16:06, 80.54s/it]
97%|█████████▋| 419/430 [9:29:59<14:45, 80.51s/it]
98%|█████████▊| 420/430 [9:31:20<13:26, 80.65s/it]
{'loss': 0.2373, 'learning_rate': 4.8692444262583224e-05, 'epoch': 9.76}
98%|█████████▊| 420/430 [9:31:20<13:26, 80.65s/it]
98%|█████████▊| 421/430 [9:32:40<12:05, 80.59s/it]
98%|█████████▊| 422/430 [9:34:01<10:44, 80.56s/it]
98%|█████████▊| 423/430 [9:35:21<09:23, 80.53s/it]
99%|█████████▊| 424/430 [9:36:42<08:03, 80.54s/it]
99%|█████████▉| 425/430 [9:38:02<06:42, 80.52s/it]
{'loss': 0.2356, 'learning_rate': 4.8786545447870833e-05, 'epoch': 9.87}
99%|█████████▉| 425/430 [9:38:02<06:42, 80.52s/it]
99%|█████████▉| 426/430 [9:39:23<05:22, 80.52s/it]
99%|█████████▉| 427/430 [9:40:43<04:01, 80.52s/it]
100%|█████████▉| 428/430 [9:42:04<02:41, 80.53s/it]
100%|█████████▉| 429/430 [9:43:24<01:20, 80.55s/it]
100%|██████████| 430/430 [9:44:45<00:00, 80.58s/it]
{'loss': 0.2362, 'learning_rate': 4.8879558733809264e-05, 'epoch': 9.99}
100%|██████████| 430/430 [9:44:45<00:00, 80.58s/it]
Training completed. Do not forget to share your model on huggingface.co/models =)
Time: 35085.85Time: 35085.80
Time: 35085.78
Samples/second: 33.41Time: 35085.88
Samples/second: 33.41
Samples/second: 33.41
Samples/second: 33.41
{'train_runtime': 35085.4942, 'train_samples_per_second': 33.413, 'train_steps_per_second': 0.012, 'train_loss': 0.412373732966046, 'epoch': 9.99}
Time: 35085.79
Time: 35085.76Samples/second: 33.41
Samples/second: 33.41
100%|██████████| 430/430 [9:44:45<00:00, 80.58s/it]GPU memory occupied: 43825 MB.
GPU memory occupied: 43825 MB.
GPU memory occupied: 43825 MB.
GPU memory occupied: 43825 MB.
GPU memory occupied: 43825 MB.
GPU memory occupied: 43825 MB.
100%|██████████| 430/430 [9:44:45<00:00, 81.59s/it]
Time: 35085.49
Samples/second: 33.41
GPU memory occupied: 43825 MB.
Configuration saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/final_checkpoint/config.json
Model weights saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/final_checkpoint/pytorch_model.bin
tokenizer config file saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/final_checkpoint/tokenizer/tokenizer_config.json
Special tokens file saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/final_checkpoint/tokenizer/special_tokens_map.json
Saving model checkpoint to experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/trainer_final_checkpoint
Configuration saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/trainer_final_checkpoint/config.json
Model weights saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/trainer_final_checkpoint/pytorch_model.bin
tokenizer config file saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/trainer_final_checkpoint/tokenizer_config.json
Special tokens file saved in experiments/2023-02-21-b0010c97cb1f06debca911602ea05b6ff85a8270fb9487d27b3d52eb4eb29e9e/trainer_final_checkpoint/special_tokens_map.json
Traceback (most recent call last):
File "tune_gpt.py", line 227, in <module>
trainer.save_state(trainer_save_dir)
TypeError: save_state() takes 1 positional argument but 2 were given
[2023-02-21 02:51:15,357] [INFO] [launch.py:350:main] Process 31459 exits successfully.
[2023-02-21 02:51:15,358] [INFO] [launch.py:350:main] Process 31463 exits successfully.
[2023-02-21 02:51:16,360] [INFO] [launch.py:350:main] Process 31486 exits successfully.
[2023-02-21 02:51:16,360] [INFO] [launch.py:350:main] Process 31471 exits successfully.
[2023-02-21 02:51:16,360] [INFO] [launch.py:350:main] Process 31478 exits successfully.
[2023-02-21 02:51:16,361] [INFO] [launch.py:350:main] Process 31490 exits successfully.
[2023-02-21 02:51:17,362] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31458
[2023-02-21 02:51:17,363] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31459
[2023-02-21 02:51:17,363] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31463
[2023-02-21 02:51:17,363] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31471
[2023-02-21 02:51:17,363] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31478
[2023-02-21 02:51:17,363] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31486
[2023-02-21 02:51:17,363] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 31490
[2023-02-21 02:51:17,364] [ERROR] [launch.py:324:sigkill_handler] ['/opt/conda/bin/python3', '-u', 'tune_gpt.py', '--local_rank=6', '--deepspeed', 'deepspeed.json', '--upload-experiment'] exits with return code = 1
/opt/conda/lib/python3.8/site-packages/xgboost/compat.py:36: FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.
from pandas import MultiIndex, Int64Index
|