File size: 142,313 Bytes
83998ab | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 | [2025-04-16 15:50:40,555] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:42,536] [WARNING] [runner.py:215:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only.
Detected VISIBLE_DEVICES=0,1,2,3,4,5,6,7: setting --include=localhost:0,1,2,3,4,5,6,7
[2025-04-16 15:50:42,536] [INFO] [runner.py:605:main] cmd = /usr/bin/python3 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29500 --enable_each_rank_log=None train.py --deepspeed scripts/newzero3.json --seed 42 --model_name_or_path /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct --train_tokenized_file /home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl --output_dir /home/stern/GRPO/offline_rl_v2/output --per_device_train_batch_size 1 --gradient_accumulation_steps 4 --evaluation_strategy no --save_strategy no --learning_rate 2e-6 --lr_scheduler_type cosine --save_only_model True --remove_unused_columns False --warmup_ratio 0.03 --num_train_epochs 3 --logging_steps 1 --report_to tensorboard --gradient_checkpointing True --overwrite_output_dir --bf16 True
[2025-04-16 15:50:44,141] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:46,067] [INFO] [launch.py:146:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}
[2025-04-16 15:50:46,067] [INFO] [launch.py:152:main] nnodes=1, num_local_procs=8, node_rank=0
[2025-04-16 15:50:46,067] [INFO] [launch.py:163:main] global_rank_mapping=defaultdict(<class 'list'>, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]})
[2025-04-16 15:50:46,067] [INFO] [launch.py:164:main] dist_world_size=8
[2025-04-16 15:50:46,067] [INFO] [launch.py:168:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
[2025-04-16 15:50:46,068] [INFO] [launch.py:256:main] process 1558860 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=0', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,068] [INFO] [launch.py:256:main] process 1558861 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=1', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,068] [INFO] [launch.py:256:main] process 1558862 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=2', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,069] [INFO] [launch.py:256:main] process 1558863 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=3', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,069] [INFO] [launch.py:256:main] process 1558864 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=4', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,069] [INFO] [launch.py:256:main] process 1558865 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=5', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,070] [INFO] [launch.py:256:main] process 1558866 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=6', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:46,070] [INFO] [launch.py:256:main] process 1558867 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=7', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_pos_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '3', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-16 15:50:49,883] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:50,768] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:50,889] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:50,972] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:51,067] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:51,070] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:51,079] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-16 15:50:51,084] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
[2025-04-16 15:50:52,644] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
[2025-04-16 15:50:52,829] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
[2025-04-16 15:50:52,955] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
[2025-04-16 15:50:53,100] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
[2025-04-16 15:50:53,195] [INFO] [comm.py:658:init_distributed] cdb=None
[2025-04-16 15:50:53,195] [INFO] [comm.py:658:init_distributed] cdb=None
[2025-04-16 15:50:53,195] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
warnings.warn(
[2025-04-16 15:50:53,238] [INFO] [comm.py:658:init_distributed] cdb=None
[2025-04-16 15:50:53,238] [INFO] [comm.py:689:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl
WARNING:__main__:Process rank: 5, device: cuda:5, n_gpu: 1
[2025-04-16 15:50:54,466] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:54,468 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
WARNING:__main__:Process rank: 0, device: cuda:0, n_gpu: 1
INFO:__main__:Training parameters CustomTrainingArguments(
_n_gpu=1,
accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False},
adafactor=False,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
auto_find_batch_size=False,
average_tokens_across_devices=False,
batch_eval_metrics=False,
bf16=True,
bf16_full_eval=False,
data_seed=None,
dataloader_drop_last=False,
dataloader_num_workers=0,
dataloader_persistent_workers=False,
dataloader_pin_memory=True,
dataloader_prefetch_factor=None,
ddp_backend=None,
ddp_broadcast_buffers=None,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
ddp_timeout=1800,
debug=[],
deepspeed=scripts/newzero3.json,
disable_tqdm=False,
dispatch_batches=None,
do_eval=False,
do_predict=False,
do_train=False,
eval_accumulation_steps=None,
eval_delay=0,
eval_do_concat_batches=True,
eval_on_start=False,
eval_steps=None,
eval_strategy=no,
eval_use_gather_object=False,
evaluation_strategy=no,
fp16=False,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
fsdp=[],
fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False},
fsdp_min_num_params=0,
fsdp_transformer_layer_cls_to_wrap=None,
full_determinism=False,
gradient_accumulation_steps=4,
gradient_checkpointing=True,
gradient_checkpointing_kwargs=None,
greater_is_better=None,
group_by_length=False,
half_precision_backend=auto,
hub_always_push=False,
hub_model_id=None,
hub_private_repo=None,
hub_strategy=every_save,
hub_token=<HUB_TOKEN>,
ignore_data_skip=False,
include_for_metrics=[],
include_inputs_for_metrics=False,
include_num_input_tokens_seen=False,
include_tokens_per_second=False,
jit_mode_eval=False,
kl_coeff=0.0,
label_names=None,
label_smoothing_factor=0.0,
learning_rate=2e-06,
length_column_name=length,
load_best_model_at_end=False,
local_rank=0,
log_level=passive,
log_level_replica=warning,
log_on_each_node=True,
logging_dir=/home/stern/GRPO/offline_rl_v2/output/runs/Apr16_15-50-53_nacamontrealdc1-p2r203n1.enovum.hivecloud.com,
logging_first_step=False,
logging_nan_inf_filter=True,
logging_steps=1.0,
logging_strategy=steps,
lr_scheduler_kwargs={},
lr_scheduler_type=cosine,
max_grad_norm=1.0,
max_steps=-1,
metric_for_best_model=None,
mp_parameters=,
neftune_noise_alpha=None,
no_cuda=False,
num_train_epochs=3.0,
optim=adamw_torch,
optim_args=None,
optim_target_modules=None,
output_dir=/home/stern/GRPO/offline_rl_v2/output,
overwrite_output_dir=True,
past_index=-1,
per_device_eval_batch_size=8,
per_device_train_batch_size=1,
prediction_loss_only=False,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
ray_scope=last,
remove_unused_columns=False,
report_to=['tensorboard'],
restore_callback_states_from_checkpoint=False,
resume_from_checkpoint=None,
run_name=/home/stern/GRPO/offline_rl_v2/output,
save_on_each_node=False,
save_only_model=True,
save_safetensors=True,
save_steps=500,
save_strategy=no,
save_total_limit=None,
seed=42,
skip_memory_metrics=True,
split_batches=None,
tf32=None,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
torch_empty_cache_steps=None,
torchdynamo=None,
tp_size=0,
tpu_metrics_debug=False,
tpu_num_cores=None,
use_cpu=False,
use_ipex=False,
use_legacy_prediction_loop=False,
use_liger_kernel=False,
use_mps_device=False,
warmup_ratio=0.03,
warmup_steps=0,
weight_decay=0.0,
)
WARNING:__main__:Process rank: 6, device: cuda:6, n_gpu: 1
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file vocab.json
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file merges.txt
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file tokenizer.json
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file added_tokens.json
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file special_tokens_map.json
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file tokenizer_config.json
[INFO|tokenization_utils_base.py:2058] 2025-04-16 15:50:54,846 >> loading file chat_template.jinja
WARNING:__main__:Process rank: 3, device: cuda:3, n_gpu: 1
WARNING:__main__:Process rank: 7, device: cuda:7, n_gpu: 1
WARNING:__main__:Process rank: 1, device: cuda:1, n_gpu: 1
WARNING:__main__:Process rank: 2, device: cuda:2, n_gpu: 1
WARNING:__main__:Process rank: 4, device: cuda:4, n_gpu: 1
[INFO|tokenization_utils_base.py:2323] 2025-04-16 15:50:55,124 >> Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
[INFO|configuration_utils.py:697] 2025-04-16 15:50:55,124 >> loading configuration file /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct/config.json
[INFO|configuration_utils.py:771] 2025-04-16 15:50:55,125 >> Model config Qwen2Config {
"architectures": [ "Qwen2ForCausalLM" ],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151645,
"hidden_act": "silu",
"hidden_size": 5120,
"initializer_range": 0.02,
"intermediate_size": 27648,
"max_position_embeddings": 32768,
"max_window_layers": 70,
"model_type": "qwen2",
"num_attention_heads": 40,
"num_hidden_layers": 64,
"num_key_value_heads": 8,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": 131072,
"tie_word_embeddings": false,
"torch_dtype": "bfloat16",
"transformers_version": "4.50.3",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 152064
}
[2025-04-16 15:50:55,134] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,136 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[INFO|modeling_utils.py:1151] 2025-04-16 15:50:55,154 >> loading weights file /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct/model.safetensors.index.json
[INFO|modeling_utils.py:1225] 2025-04-16 15:50:55,155 >> Will use torch_dtype=torch.bfloat16 as defined in model's config object
[INFO|modeling_utils.py:2170] 2025-04-16 15:50:55,155 >> Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
[INFO|modeling_utils.py:3747] 2025-04-16 15:50:55,155 >> Detected DeepSpeed ZeRO-3: activating zero.init() for this model
[2025-04-16 15:50:55,155] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,157 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[INFO|configuration_utils.py:1139] 2025-04-16 15:50:55,161 >> Generate config GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": 151645
}
[2025-04-16 15:50:55,291] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,293 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-16 15:50:55,362] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,365 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-16 15:50:55,409] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,411 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-16 15:50:55,426] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,429 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-16 15:50:55,440] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-16 15:50:55,443 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-16 15:51:17,163] [INFO] [partition_parameters.py:348:__exit__] finished initializing model - num_params = 771, num_elems = 32.76B
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:11, 1.44it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:10, 1.57it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:10, 1.56it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:11, 1.45it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:10, 1.49it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:10, 1.50it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:11, 1.41it/s]
Loading checkpoint shards: 6%|▌ | 1/17 [00:00<00:13, 1.23it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.28it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.32it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.31it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.29it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.30it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.28it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:11, 1.32it/s]
Loading checkpoint shards: 12%|█▏ | 2/17 [00:01<00:12, 1.17it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.20it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.20it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.20it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.21it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.21it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.22it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:11, 1.22it/s]
Loading checkpoint shards: 18%|█▊ | 3/17 [00:02<00:12, 1.14it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.22it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.22it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.23it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.23it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.23it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.22it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.23it/s]
Loading checkpoint shards: 24%|██▎ | 4/17 [00:03<00:10, 1.21it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.27it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.26it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.26it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.26it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.26it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.26it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:03<00:09, 1.27it/s]
Loading checkpoint shards: 29%|██▉ | 5/17 [00:04<00:09, 1.26it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.31it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.31it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.30it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.30it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.30it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.30it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.30it/s]
Loading checkpoint shards: 35%|███▌ | 6/17 [00:04<00:08, 1.30it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 41%|████ | 7/17 [00:05<00:07, 1.33it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 47%|████▋ | 8/17 [00:06<00:06, 1.35it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 53%|█████▎ | 9/17 [00:06<00:05, 1.36it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.37it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.37it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.37it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.37it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.36it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.37it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.37it/s]
Loading checkpoint shards: 59%|█████▉ | 10/17 [00:07<00:05, 1.36it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.36it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.35it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.35it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.35it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.36it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.35it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.35it/s]
Loading checkpoint shards: 65%|██████▍ | 11/17 [00:08<00:04, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.34it/s]
Loading checkpoint shards: 71%|███████ | 12/17 [00:09<00:03, 1.33it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 76%|███████▋ | 13/17 [00:09<00:02, 1.34it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.35it/s]
Loading checkpoint shards: 82%|████████▏ | 14/17 [00:10<00:02, 1.34it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 88%|████████▊ | 15/17 [00:11<00:01, 1.35it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:12<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:11<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:11<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:12<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:12<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:12<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:12<00:00, 1.36it/s]
Loading checkpoint shards: 94%|█████████▍| 16/17 [00:12<00:00, 1.37it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.35it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.45it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00, 1.34it/s]
[INFO|modeling_utils.py:4987] 2025-04-16 15:51:29,942 >> All model checkpoint weights were used when initializing Qwen2ForCausalLM.
[INFO|modeling_utils.py:4995] 2025-04-16 15:51:29,942 >> All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
[INFO|configuration_utils.py:1092] 2025-04-16 15:51:29,947 >> loading configuration file /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct/generation_config.json
[INFO|configuration_utils.py:1139] 2025-04-16 15:51:29,947 >> Generate config GenerationConfig {
"bos_token_id": 151643,
"do_sample": true,
"eos_token_id": [
151645,
151643
],
"pad_token_id": 151643,
"repetition_penalty": 1.05,
"temperature": 0.7,
"top_k": 20,
"top_p": 0.8
}
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
Using custom data configuration default-58a08ad63b752e90
INFO:datasets.builder:Using custom data configuration default-58a08ad63b752e90
Loading Dataset Infos from /home/stern/.local/lib/python3.10/site-packages/datasets/packaged_modules/json
INFO:datasets.info:Loading Dataset Infos from /home/stern/.local/lib/python3.10/site-packages/datasets/packaged_modules/json
Overwrite dataset info from restored data version if exists.
INFO:datasets.builder:Overwrite dataset info from restored data version if exists.
Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-58a08ad63b752e90/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
INFO:datasets.info:Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-58a08ad63b752e90/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
Found cached dataset json (/home/stern/.cache/huggingface/datasets/json/default-58a08ad63b752e90/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092)
INFO:datasets.builder:Found cached dataset json (/home/stern/.cache/huggingface/datasets/json/default-58a08ad63b752e90/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092)
Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-58a08ad63b752e90/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
INFO:datasets.info:Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-58a08ad63b752e90/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
trainer = OfflineREINFORCETrainer(
[INFO|trainer.py:748] 2025-04-16 15:51:30,254 >> Using auto half precision backend
INFO:__main__:*** Train ***
[INFO|deepspeed.py:386] 2025-04-16 15:51:30,446 >> Detected ZeRO Offload and non-DeepSpeed optimizers: This combination should work as long as the custom optimizer has both CPU and GPU implementation (except LAMB)
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/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)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.362370252609253 seconds
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/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)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.385770320892334 seconds
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/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)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.2810471057891846 seconds
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/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)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.278972864151001 seconds
Adam Optimizer #0 is created with AVX512 arithmetic capability.
Config: alpha=0.000002, betas=(0.900000, 0.999000), weight_decay=0.010000, adam_w=1
[2025-04-16 15:51:34,212] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed info: version=0.16.5, git-hash=unknown, git-branch=unknown
[2025-04-16 15:51:34,212] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[2025-04-16 15:51:34,229] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False
[2025-04-16 15:51:34,231] [INFO] [logging.py:107:log_dist] [Rank 0] Using client Optimizer as basic optimizer
[2025-04-16 15:51:34,231] [INFO] [logging.py:107:log_dist] [Rank 0] Removing param_group that has no 'params' in the basic Optimizer
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.4311106204986572 seconds
[2025-04-16 15:51:34,275] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed Basic Optimizer = DeepSpeedCPUAdam
[2025-04-16 15:51:34,275] [INFO] [utils.py:59:is_zero_supported_optimizer] Checking ZeRO support for optimizer=DeepSpeedCPUAdam type=<class 'deepspeed.ops.adam.cpu_adam.DeepSpeedCPUAdam'>
[2025-04-16 15:51:34,275] [INFO] [logging.py:107:log_dist] [Rank 0] Creating fp16 ZeRO stage 3 optimizer, MiCS is enabled False, Hierarchical params gather False
[2025-04-16 15:51:34,275] [INFO] [logging.py:107:log_dist] [Rank 0] Creating torch.bfloat16 ZeRO stage 3 optimizer
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/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)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.335779905319214 seconds
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/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)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.413490056991577 seconds
[2025-04-16 15:51:34,405] [INFO] [utils.py:781:see_memory_usage] Stage 3 initialize beginning
[2025-04-16 15:51:34,405] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 2.9 GB CA 0.0 GB Max_CA 3 GB
[2025-04-16 15:51:34,406] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 137.42 GB, percent = 13.6%
[2025-04-16 15:51:34,408] [INFO] [stage3.py:170:__init__] Reduce bucket size 100000000
[2025-04-16 15:51:34,408] [INFO] [stage3.py:171:__init__] Prefetch bucket size 100000000
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.5096564292907715 seconds
[2025-04-16 15:51:34,506] [INFO] [utils.py:781:see_memory_usage] DeepSpeedZeRoOffload initialize [begin]
[2025-04-16 15:51:34,507] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:51:34,507] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 137.44 GB, percent = 13.6%
Parameter Offload: Total persistent parameters: 1119232 in 321 params
[2025-04-16 15:51:34,672] [INFO] [utils.py:781:see_memory_usage] DeepSpeedZeRoOffload initialize [end]
[2025-04-16 15:51:34,672] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:51:34,672] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 137.47 GB, percent = 13.6%
[2025-04-16 15:51:34,779] [INFO] [utils.py:781:see_memory_usage] Before creating fp16 partitions
[2025-04-16 15:51:34,780] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:51:34,780] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 137.47 GB, percent = 13.6%
[2025-04-16 15:51:57,066] [INFO] [utils.py:781:see_memory_usage] After creating fp16 partitions: 41
[2025-04-16 15:51:57,067] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:51:57,067] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 236.86 GB, percent = 23.5%
[2025-04-16 15:51:57,421] [INFO] [utils.py:781:see_memory_usage] Before creating fp32 partitions
[2025-04-16 15:51:57,422] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:51:57,422] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 245.91 GB, percent = 24.4%
[2025-04-16 15:52:00,950] [INFO] [utils.py:781:see_memory_usage] After creating fp32 partitions
[2025-04-16 15:52:00,950] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:52:00,950] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 329.28 GB, percent = 32.7%
[2025-04-16 15:52:01,139] [INFO] [utils.py:781:see_memory_usage] Before initializing optimizer states
[2025-04-16 15:52:01,140] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:52:01,140] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 334.99 GB, percent = 33.3%
[2025-04-16 15:52:13,760] [INFO] [utils.py:781:see_memory_usage] After initializing optimizer states
[2025-04-16 15:52:13,761] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB Max_MA 0.0 GB CA 0.0 GB Max_CA 0 GB
[2025-04-16 15:52:13,761] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 511.84 GB, percent = 50.8%
[2025-04-16 15:52:14,279] [INFO] [stage3.py:534:_setup_for_real_optimizer] optimizer state initialized
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
[WARNING|logging.py:329] 2025-04-16 15:52:20,609 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-16 15:52:20,611 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-16 15:52:20,612 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-16 15:52:20,615 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-16 15:52:20,615 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-16 15:52:20,618 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-16 15:52:20,619 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[2025-04-16 15:52:20,720] [INFO] [utils.py:781:see_memory_usage] After initializing ZeRO optimizer
[2025-04-16 15:52:20,721] [INFO] [utils.py:782:see_memory_usage] MA 0.19 GB Max_MA 3.09 GB CA 3.09 GB Max_CA 3 GB
[2025-04-16 15:52:20,721] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory: used = 563.25 GB, percent = 55.9%
[2025-04-16 15:52:20,721] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed Final Optimizer = DeepSpeedZeroOptimizer_Stage3
[2025-04-16 15:52:20,721] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed using configured LR scheduler = None
[2025-04-16 15:52:20,721] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed LR Scheduler = None
[2025-04-16 15:52:20,721] [INFO] [logging.py:107:log_dist] [Rank 0] step=0, skipped=0, lr=[0.0, 0.0], mom=[(0.9, 0.999), (0.9, 0.999)]
[2025-04-16 15:52:20,723] [INFO] [config.py:1000:print] DeepSpeedEngine configuration:
[2025-04-16 15:52:20,723] [INFO] [config.py:1004:print] activation_checkpointing_config {
"partition_activations": false,
"contiguous_memory_optimization": false,
"cpu_checkpointing": false,
"number_checkpoints": null,
"synchronize_checkpoint_boundary": false,
"profile": false
}
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'intra_op_parallelism': 1, 'single_submit': False, 'overlap_events': True, 'use_gds': False}
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] amp_enabled .................. False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] amp_params ................... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004: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
}
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] bfloat16_enabled ............. True
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] bfloat16_immediate_grad_update True
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] checkpoint_parallel_write_pipeline False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] checkpoint_tag_validation_enabled True
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] checkpoint_tag_validation_fail False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] comms_config ................. <deepspeed.comm.config.DeepSpeedCommsConfig object at 0x7a7bbbd938b0>
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] communication_data_type ...... None
[2025-04-16 15:52:20,724] [INFO] [config.py:1004: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}}
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] curriculum_enabled_legacy .... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] curriculum_params_legacy ..... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] data_efficiency_config ....... {'enabled': False, 'seed': 1234, 'data_sampling': {'enabled': False, 'num_epochs': 1000, 'num_workers': 0, 'pin_memory': False, 'curriculum_learning': {'enabled': False}, 'dynamic_batching': {'enabled': False, 'lr_scaling_method': 'linear', 'min_batch_size': 1, 'max_batch_size': None, 'sequence_picking_order': 'dataloader', 'verbose': False}}, 'data_routing': {'enabled': False, 'random_ltd': {'enabled': False, 'layer_token_lr_schedule': {'enabled': False}}}}
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] data_efficiency_enabled ...... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] dataloader_drop_last ......... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] disable_allgather ............ False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] dump_state ................... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] dynamic_loss_scale_args ...... None
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_enabled ........... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_gas_boundary_resolution 1
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_layer_name ........ bert.encoder.layer
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_layer_num ......... 0
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_max_iter .......... 100
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_stability ......... 1e-06
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_tol ............... 0.01
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] eigenvalue_verbose ........... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] elasticity_enabled ........... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004: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
}
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] fp16_auto_cast ............... None
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] fp16_enabled ................. False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] fp16_master_weights_and_gradients False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] global_rank .................. 0
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] grad_accum_dtype ............. None
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] gradient_accumulation_steps .. 4
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] gradient_clipping ............ 1.0
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] gradient_predivide_factor .... 1.0
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] graph_harvesting ............. False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004: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
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] initial_dynamic_scale ........ 1
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] load_universal_checkpoint .... False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] loss_scale ................... 1.0
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] memory_breakdown ............. False
[2025-04-16 15:52:20,724] [INFO] [config.py:1004:print] mics_hierarchial_params_gather False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] mics_shard_size .............. -1
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] monitor_config ............... tensorboard=TensorBoardConfig(enabled=False, output_path='', job_name='DeepSpeedJobName') comet=CometConfig(enabled=False, samples_log_interval=100, project=None, workspace=None, api_key=None, experiment_name=None, experiment_key=None, online=None, mode=None) wandb=WandbConfig(enabled=False, group=None, team=None, project='deepspeed') csv_monitor=CSVConfig(enabled=False, output_path='', job_name='DeepSpeedJobName')
[2025-04-16 15:52:20,725] [INFO] [config.py:1004: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
}
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] optimizer_legacy_fusion ...... False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] optimizer_name ............... None
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] optimizer_params ............. None
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] pipeline ..................... {'stages': 'auto', 'partition': 'best', 'seed_layers': False, 'activation_checkpoint_interval': 0, 'pipe_partitioned': True, 'grad_partitioned': True}
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] pld_enabled .................. False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] pld_params ................... False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] prescale_gradients ........... False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] scheduler_name ............... None
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] scheduler_params ............. None
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] seq_parallel_communication_data_type torch.float32
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] sparse_attention ............. None
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] sparse_gradients_enabled ..... False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] steps_per_print .............. inf
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] tensor_parallel_config ....... dtype=torch.float16 autotp_size=0 tensor_parallel=TPConfig(tp_size=1, tp_grain_size=1, mpu=None, tp_group=None) injection_policy_tuple=None keep_module_on_host=False replace_with_kernel_inject=False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] timers_config ................ enabled=True synchronized=True
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] train_batch_size ............. 32
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] train_micro_batch_size_per_gpu 1
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] use_data_before_expert_parallel_ False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] use_node_local_storage ....... False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] wall_clock_breakdown ......... False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] weight_quantization_config ... None
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] world_size ................... 8
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] zero_allow_untested_optimizer True
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] zero_config .................. stage=3 contiguous_gradients=True reduce_scatter=True reduce_bucket_size=100000000 use_multi_rank_bucket_allreduce=True allgather_partitions=True allgather_bucket_size=500000000 overlap_comm=True load_from_fp32_weights=True elastic_checkpoint=False offload_param=DeepSpeedZeroOffloadParamConfig(device='cpu', nvme_path=None, buffer_count=5, buffer_size=100000000, max_in_cpu=1000000000, pin_memory=True) offload_optimizer=DeepSpeedZeroOffloadOptimizerConfig(device='cpu', nvme_path=None, buffer_count=4, pin_memory=True, pipeline_read=False, pipeline_write=False, fast_init=False, ratio=1.0) sub_group_size=100000000 cpu_offload_param=None cpu_offload_use_pin_memory=None cpu_offload=None prefetch_bucket_size=100000000 param_persistence_threshold=100000 model_persistence_threshold=9223372036854775807 max_live_parameters=100000000 max_reuse_distance=100000000 gather_16bit_weights_on_model_save=True module_granularity_threshold=0 use_all_reduce_for_fetch_params=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 zeropp_loco_param=None mics_shard_size=-1 mics_hierarchical_params_gather=False memory_efficient_linear=True pipeline_loading_checkpoint=False override_module_apply=True log_trace_cache_warnings=False
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] zero_enabled ................. True
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] zero_force_ds_cpu_optimizer .. True
[2025-04-16 15:52:20,725] [INFO] [config.py:1004:print] zero_optimization_stage ...... 3
[2025-04-16 15:52:20,725] [INFO] [config.py:990:print_user_config] json = {
"fp16": {
"enabled": false
},
"bf16": {
"enabled": true
},
"train_micro_batch_size_per_gpu": 1,
"gradient_accumulation_steps": 4,
"zero_optimization": {
"stage": 3,
"offload_optimizer": {
"device": "cpu",
"pin_memory": true
},
"offload_param": {
"device": "cpu",
"pin_memory": true
},
"overlap_comm": true,
"contiguous_gradients": true,
"sub_group_size": 1.000000e+08,
"reduce_bucket_size": 1.000000e+08,
"stage3_prefetch_bucket_size": 1.000000e+08,
"stage3_param_persistence_threshold": 1.000000e+05,
"stage3_max_live_parameters": 1.000000e+08,
"stage3_max_reuse_distance": 1.000000e+08,
"stage3_gather_16bit_weights_on_model_save": true
},
"gradient_clipping": 1.0,
"wall_clock_breakdown": false,
"steps_per_print": inf,
"zero_allow_untested_optimizer": true
}
[INFO|trainer.py:2409] 2025-04-16 15:52:20,725 >> ***** Running training *****
[INFO|trainer.py:2410] 2025-04-16 15:52:20,725 >> Num examples = 2,245
[INFO|trainer.py:2411] 2025-04-16 15:52:20,725 >> Num Epochs = 3
[INFO|trainer.py:2412] 2025-04-16 15:52:20,725 >> Instantaneous batch size per device = 1
[INFO|trainer.py:2415] 2025-04-16 15:52:20,725 >> Total train batch size (w. parallel, distributed & accumulation) = 32
[INFO|trainer.py:2416] 2025-04-16 15:52:20,725 >> Gradient Accumulation steps = 4
[INFO|trainer.py:2417] 2025-04-16 15:52:20,725 >> Total optimization steps = 210
[INFO|trainer.py:2418] 2025-04-16 15:52:20,727 >> Number of trainable parameters = 32,763,876,352
0%| | 0/210 [00:00<?, ?it/s]/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
[WARNING|logging.py:329] 2025-04-16 15:52:20,791 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs): # type: ignore[attr-defined]
0%| | 1/210 [00:27<1:36:37, 27.74s/it]
{'loss': 0.0662, 'grad_norm': 1.1819676160812378, 'learning_rate': 2.857142857142857e-07, 'kl': -0.0002, 'entropy': 0.0107, 'ce_loss': 0.013, 'epoch': 0.01}
0%| | 1/210 [00:27<1:36:37, 27.74s/it]
1%| | 2/210 [00:45<1:15:08, 21.67s/it]
{'loss': 0.0673, 'grad_norm': 1.3304132223129272, 'learning_rate': 5.714285714285714e-07, 'kl': 0.0003, 'entropy': 0.0011, 'ce_loss': 0.0261, 'epoch': 0.03}
1%| | 2/210 [00:45<1:15:08, 21.67s/it]
1%|▏ | 3/210 [01:02<1:08:16, 19.79s/it]
{'loss': 0.0606, 'grad_norm': 1.2711843252182007, 'learning_rate': 8.57142857142857e-07, 'kl': -0.0001, 'entropy': 0.0014, 'ce_loss': 0.0125, 'epoch': 0.04}
1%|▏ | 3/210 [01:02<1:08:16, 19.79s/it]
2%|▏ | 4/210 [01:20<1:04:48, 18.87s/it]
{'loss': 0.0652, 'grad_norm': 1.2260750532150269, 'learning_rate': 1.1428571428571428e-06, 'kl': 0.0005, 'entropy': -0.0415, 'ce_loss': 0.0083, 'epoch': 0.06}
2%|▏ | 4/210 [01:20<1:04:48, 18.87s/it]
2%|▏ | 5/210 [01:37<1:02:29, 18.29s/it]
{'loss': 0.0619, 'grad_norm': 0.9697835445404053, 'learning_rate': 1.4285714285714286e-06, 'kl': 0.0008, 'entropy': 0.0374, 'ce_loss': 0.0204, 'epoch': 0.07}
2%|▏ | 5/210 [01:37<1:02:29, 18.29s/it]
3%|▎ | 6/210 [01:54<1:01:02, 17.95s/it]
{'loss': 0.0623, 'grad_norm': 0.9166115522384644, 'learning_rate': 1.714285714285714e-06, 'kl': 0.0023, 'entropy': -0.0053, 'ce_loss': 0.0114, 'epoch': 0.09}
3%|▎ | 6/210 [01:54<1:01:02, 17.95s/it]
3%|▎ | 7/210 [02:12<1:00:01, 17.74s/it]
{'loss': 0.0518, 'grad_norm': 0.381746768951416, 'learning_rate': 2e-06, 'kl': 0.0104, 'entropy': -0.0957, 'ce_loss': 0.0097, 'epoch': 0.1}
3%|▎ | 7/210 [02:12<1:00:01, 17.74s/it]
4%|▍ | 8/210 [02:29<59:11, 17.58s/it]
{'loss': 0.0597, 'grad_norm': 0.44386395812034607, 'learning_rate': 1.999880251796685e-06, 'kl': 0.0067, 'entropy': -0.0781, 'ce_loss': 0.0039, 'epoch': 0.11}
4%|▍ | 8/210 [02:29<59:11, 17.58s/it]
4%|▍ | 9/210 [02:46<58:22, 17.43s/it]
{'loss': 0.0591, 'grad_norm': 0.43770262598991394, 'learning_rate': 1.9995210358660034e-06, 'kl': 0.0077, 'entropy': -0.063, 'ce_loss': 0.0192, 'epoch': 0.13}
4%|▍ | 9/210 [02:46<58:22, 17.43s/it]
5%|▍ | 10/210 [03:03<57:53, 17.37s/it]
{'loss': 0.0529, 'grad_norm': 0.4576289653778076, 'learning_rate': 1.998922438238881e-06, 'kl': 0.0027, 'entropy': -0.0292, 'ce_loss': 0.0108, 'epoch': 0.14}
5%|▍ | 10/210 [03:03<57:53, 17.37s/it]
5%|▌ | 11/210 [03:21<57:51, 17.45s/it]
{'loss': 0.0441, 'grad_norm': 0.3885749280452728, 'learning_rate': 1.9980846022772976e-06, 'kl': 0.0132, 'entropy': -0.0564, 'ce_loss': 0.004, 'epoch': 0.16}
5%|▌ | 11/210 [03:21<57:51, 17.45s/it]
6%|▌ | 12/210 [03:38<57:15, 17.35s/it]
{'loss': 0.0541, 'grad_norm': 0.4447019398212433, 'learning_rate': 1.9970077286399556e-06, 'kl': 0.0209, 'entropy': -0.085, 'ce_loss': 0.0159, 'epoch': 0.17}
6%|▌ | 12/210 [03:38<57:15, 17.35s/it]
6%|▌ | 13/210 [03:55<56:38, 17.25s/it]
{'loss': 0.0582, 'grad_norm': 0.5154215097427368, 'learning_rate': 1.9956920752342224e-06, 'kl': 0.0087, 'entropy': -0.0281, 'ce_loss': 0.0058, 'epoch': 0.19}
6%|▌ | 13/210 [03:55<56:38, 17.25s/it]
7%|▋ | 14/210 [04:12<56:12, 17.21s/it]
{'loss': 0.0568, 'grad_norm': 0.4587293565273285, 'learning_rate': 1.9941379571543595e-06, 'kl': 0.0067, 'entropy': -0.0574, 'ce_loss': 0.0104, 'epoch': 0.2}
7%|▋ | 14/210 [04:12<56:12, 17.21s/it]
7%|▋ | 15/210 [04:29<55:55, 17.21s/it]
{'loss': 0.0471, 'grad_norm': 0.4319175183773041, 'learning_rate': 1.992345746606063e-06, 'kl': 0.0064, 'entropy': -0.0276, 'ce_loss': 0.006, 'epoch': 0.21}
7%|▋ | 15/210 [04:29<55:55, 17.21s/it]
8%|▊ | 16/210 [04:46<55:41, 17.23s/it]
{'loss': 0.0612, 'grad_norm': 0.5637874603271484, 'learning_rate': 1.9903158728173205e-06, 'kl': 0.009, 'entropy': -0.0898, 'ce_loss': 0.0122, 'epoch': 0.23}
8%|▊ | 16/210 [04:46<55:41, 17.23s/it]
8%|▊ | 17/210 [05:04<55:27, 17.24s/it]
{'loss': 0.0494, 'grad_norm': 0.5410430431365967, 'learning_rate': 1.9880488219356083e-06, 'kl': 0.0154, 'entropy': -0.0168, 'ce_loss': 0.0177, 'epoch': 0.24}
8%|▊ | 17/210 [05:04<55:27, 17.24s/it]
9%|▊ | 18/210 [05:21<55:02, 17.20s/it]
{'loss': 0.0599, 'grad_norm': 0.5834937691688538, 'learning_rate': 1.9855451369114677e-06, 'kl': -0.0065, 'entropy': 0.0144, 'ce_loss': 0.0217, 'epoch': 0.26}
9%|▊ | 18/210 [05:21<55:02, 17.20s/it]
9%|▉ | 19/210 [05:38<54:37, 17.16s/it]
{'loss': 0.0515, 'grad_norm': 0.46324604749679565, 'learning_rate': 1.9828054173684644e-06, 'kl': 0.0063, 'entropy': -0.0138, 'ce_loss': 0.0166, 'epoch': 0.27}
9%|▉ | 19/210 [05:38<54:37, 17.16s/it]
10%|▉ | 20/210 [05:55<54:21, 17.17s/it]
{'loss': 0.0511, 'grad_norm': 0.41407546401023865, 'learning_rate': 1.9798303194595845e-06, 'kl': 0.012, 'entropy': -0.0474, 'ce_loss': 0.0168, 'epoch': 0.28}
10%|▉ | 20/210 [05:55<54:21, 17.17s/it]
10%|█ | 21/210 [06:12<54:13, 17.22s/it]
{'loss': 0.0545, 'grad_norm': 0.42467767000198364, 'learning_rate': 1.9766205557100868e-06, 'kl': 0.005, 'entropy': -0.0322, 'ce_loss': 0.0121, 'epoch': 0.3}
10%|█ | 21/210 [06:12<54:13, 17.22s/it]
10%|█ | 22/210 [06:30<53:49, 17.18s/it]
{'loss': 0.0468, 'grad_norm': 0.40804415941238403, 'learning_rate': 1.973176894846855e-06, 'kl': 0.0059, 'entropy': 0.0398, 'ce_loss': 0.0073, 'epoch': 0.31}
10%|█ | 22/210 [06:30<53:49, 17.18s/it]
11%|█ | 23/210 [06:47<53:39, 17.22s/it]
{'loss': 0.0591, 'grad_norm': 0.5296483039855957, 'learning_rate': 1.969500161614291e-06, 'kl': 0.0094, 'entropy': 0.0023, 'ce_loss': 0.0159, 'epoch': 0.33}
11%|█ | 23/210 [06:47<53:39, 17.22s/it]
11%|█▏ | 24/210 [07:04<53:36, 17.29s/it]
{'loss': 0.0512, 'grad_norm': 0.48807722330093384, 'learning_rate': 1.965591236576794e-06, 'kl': 0.009, 'entropy': 0.0148, 'ce_loss': 0.0158, 'epoch': 0.34}
11%|█▏ | 24/210 [07:04<53:36, 17.29s/it]
12%|█▏ | 25/210 [07:22<53:14, 17.27s/it]
{'loss': 0.0583, 'grad_norm': 0.4568687677383423, 'learning_rate': 1.9614510559078624e-06, 'kl': 0.0106, 'entropy': -0.0532, 'ce_loss': 0.0334, 'epoch': 0.36}
12%|█▏ | 25/210 [07:22<53:14, 17.27s/it]
12%|█▏ | 26/210 [07:39<52:42, 17.19s/it]
{'loss': 0.0561, 'grad_norm': 0.43054699897766113, 'learning_rate': 1.9570806111658896e-06, 'kl': 0.0067, 'entropy': -0.0596, 'ce_loss': 0.0152, 'epoch': 0.37}
12%|█▏ | 26/210 [07:39<52:42, 17.19s/it]
13%|█▎ | 27/210 [07:56<52:15, 17.13s/it]
{'loss': 0.0506, 'grad_norm': 0.42163872718811035, 'learning_rate': 1.9524809490566874e-06, 'kl': 0.0248, 'entropy': -0.1104, 'ce_loss': 0.0152, 'epoch': 0.38}
13%|█▎ | 27/210 [07:56<52:15, 17.13s/it]
13%|█▎ | 28/210 [08:13<51:58, 17.14s/it]
{'loss': 0.0521, 'grad_norm': 0.39460843801498413, 'learning_rate': 1.9476531711828025e-06, 'kl': 0.0066, 'entropy': -0.0137, 'ce_loss': 0.0149, 'epoch': 0.4}
13%|█▎ | 28/210 [08:13<51:58, 17.14s/it]
14%|█▍ | 29/210 [08:30<51:42, 17.14s/it]
{'loss': 0.0518, 'grad_norm': 0.3918153643608093, 'learning_rate': 1.942598433779687e-06, 'kl': 0.0017, 'entropy': -0.0325, 'ce_loss': 0.0062, 'epoch': 0.41}
14%|█▍ | 29/210 [08:30<51:42, 17.14s/it]
14%|█▍ | 30/210 [08:47<51:38, 17.21s/it]
{'loss': 0.0565, 'grad_norm': 0.40968766808509827, 'learning_rate': 1.9373179474387856e-06, 'kl': 0.0131, 'entropy': -0.0197, 'ce_loss': 0.0108, 'epoch': 0.43}
14%|█▍ | 30/210 [08:47<51:38, 17.21s/it]
15%|█▍ | 31/210 [09:04<51:15, 17.18s/it]
{'loss': 0.0527, 'grad_norm': 0.4086287021636963, 'learning_rate': 1.931812976817603e-06, 'kl': 0.0023, 'entropy': 0.0013, 'ce_loss': 0.0102, 'epoch': 0.44}
15%|█▍ | 31/210 [09:04<51:15, 17.18s/it]
15%|█▌ | 32/210 [09:21<50:50, 17.14s/it]
{'loss': 0.0566, 'grad_norm': 0.4306926727294922, 'learning_rate': 1.9260848403368207e-06, 'kl': 0.0132, 'entropy': -0.0337, 'ce_loss': 0.0139, 'epoch': 0.46}
15%|█▌ | 32/210 [09:21<50:50, 17.14s/it]
16%|█▌ | 33/210 [09:39<50:42, 17.19s/it]
{'loss': 0.0524, 'grad_norm': 0.4188339114189148, 'learning_rate': 1.920134909864543e-06, 'kl': 0.0137, 'entropy': -0.0422, 'ce_loss': 0.0135, 'epoch': 0.47}
16%|█▌ | 33/210 [09:39<50:42, 17.19s/it]
16%|█▌ | 34/210 [09:56<50:23, 17.18s/it]
{'loss': 0.0504, 'grad_norm': 0.41842740774154663, 'learning_rate': 1.9139646103877378e-06, 'kl': 0.0079, 'entropy': -0.0114, 'ce_loss': 0.007, 'epoch': 0.48}
16%|█▌ | 34/210 [09:56<50:23, 17.18s/it]
17%|█▋ | 35/210 [10:13<50:03, 17.16s/it]
{'loss': 0.049, 'grad_norm': 0.4514538049697876, 'learning_rate': 1.907575419670957e-06, 'kl': 0.0148, 'entropy': -0.0454, 'ce_loss': 0.0059, 'epoch': 0.5}
17%|█▋ | 35/210 [10:13<50:03, 17.16s/it]
17%|█▋ | 36/210 [10:30<49:44, 17.15s/it]
{'loss': 0.0452, 'grad_norm': 0.39393845200538635, 'learning_rate': 1.9009688679024189e-06, 'kl': 0.0146, 'entropy': -0.125, 'ce_loss': 0.0035, 'epoch': 0.51}
17%|█▋ | 36/210 [10:30<49:44, 17.15s/it]
18%|█▊ | 37/210 [10:47<49:37, 17.21s/it]
{'loss': 0.0542, 'grad_norm': 0.3971409201622009, 'learning_rate': 1.8941465373275328e-06, 'kl': 0.0098, 'entropy': -0.0044, 'ce_loss': 0.0084, 'epoch': 0.53}
18%|█▊ | 37/210 [10:47<49:37, 17.21s/it]
18%|█▊ | 38/210 [11:04<49:13, 17.17s/it]
{'loss': 0.0535, 'grad_norm': 0.42538702487945557, 'learning_rate': 1.8871100618699552e-06, 'kl': 0.0118, 'entropy': -0.0598, 'ce_loss': 0.0087, 'epoch': 0.54}
18%|█▊ | 38/210 [11:04<49:13, 17.17s/it]
19%|█▊ | 39/210 [11:22<49:00, 17.20s/it]
{'loss': 0.0459, 'grad_norm': 0.4148142337799072, 'learning_rate': 1.8798611267402744e-06, 'kl': 0.0255, 'entropy': -0.167, 'ce_loss': 0.0111, 'epoch': 0.56}
19%|█▊ | 39/210 [11:22<49:00, 17.20s/it]
19%|█▉ | 40/210 [11:39<48:39, 17.17s/it]
{'loss': 0.0542, 'grad_norm': 0.40929272770881653, 'learning_rate': 1.8724014680324057e-06, 'kl': 0.0051, 'entropy': -0.0879, 'ce_loss': 0.0121, 'epoch': 0.57}
19%|█▉ | 40/210 [11:39<48:39, 17.17s/it]
20%|█▉ | 41/210 [11:56<48:16, 17.14s/it]
{'loss': 0.053, 'grad_norm': 0.42668187618255615, 'learning_rate': 1.8647328723078037e-06, 'kl': 0.0089, 'entropy': -0.0227, 'ce_loss': 0.0077, 'epoch': 0.58}
20%|█▉ | 41/210 [11:56<48:16, 17.14s/it]
20%|██ | 42/210 [12:13<47:59, 17.14s/it]
{'loss': 0.0564, 'grad_norm': 0.4324919879436493, 'learning_rate': 1.8568571761675891e-06, 'kl': 0.0072, 'entropy': 0.0236, 'ce_loss': 0.0123, 'epoch': 0.6}
20%|██ | 42/210 [12:13<47:59, 17.14s/it]
20%|██ | 43/210 [12:30<47:43, 17.15s/it]
{'loss': 0.062, 'grad_norm': 0.4570630192756653, 'learning_rate': 1.8487762658126868e-06, 'kl': 0.0055, 'entropy': -0.0204, 'ce_loss': 0.0127, 'epoch': 0.61}
20%|██ | 43/210 [12:30<47:43, 17.15s/it]
21%|██ | 44/210 [12:48<47:33, 17.19s/it]
{'loss': 0.0518, 'grad_norm': 0.40118956565856934, 'learning_rate': 1.8404920765920894e-06, 'kl': 0.0078, 'entropy': -0.006, 'ce_loss': 0.0169, 'epoch': 0.63}
21%|██ | 44/210 [12:48<47:33, 17.19s/it]
21%|██▏ | 45/210 [13:05<47:18, 17.20s/it]
{'loss': 0.0511, 'grad_norm': 0.41626226902008057, 'learning_rate': 1.8320065925393467e-06, 'kl': 0.0081, 'entropy': -0.0442, 'ce_loss': 0.0087, 'epoch': 0.64}
21%|██▏ | 45/210 [13:05<47:18, 17.20s/it]
22%|██▏ | 46/210 [13:22<46:59, 17.19s/it]
{'loss': 0.0525, 'grad_norm': 0.5035837292671204, 'learning_rate': 1.8233218458973983e-06, 'kl': 0.0085, 'entropy': -0.041, 'ce_loss': 0.0108, 'epoch': 0.65}
22%|██▏ | 46/210 [13:22<46:59, 17.19s/it]
22%|██▏ | 47/210 [13:39<46:43, 17.20s/it]
{'loss': 0.055, 'grad_norm': 0.42323240637779236, 'learning_rate': 1.8144399166318568e-06, 'kl': 0.0101, 'entropy': -0.0654, 'ce_loss': 0.0162, 'epoch': 0.67}
22%|██▏ | 47/210 [13:39<46:43, 17.20s/it]
23%|██▎ | 48/210 [13:56<46:21, 17.17s/it]
{'loss': 0.0548, 'grad_norm': 0.416448175907135, 'learning_rate': 1.805362931932866e-06, 'kl': 0.0081, 'entropy': -0.0337, 'ce_loss': 0.0227, 'epoch': 0.68}
23%|██▎ | 48/210 [13:56<46:21, 17.17s/it]
23%|██▎ | 49/210 [14:13<45:58, 17.14s/it]
{'loss': 0.0602, 'grad_norm': 0.4205745756626129, 'learning_rate': 1.7960930657056437e-06, 'kl': 0.009, 'entropy': -0.0236, 'ce_loss': 0.0165, 'epoch': 0.7}
23%|██▎ | 49/210 [14:13<45:58, 17.14s/it]
24%|██▍ | 50/210 [14:31<45:50, 17.19s/it]
{'loss': 0.054, 'grad_norm': 0.47851261496543884, 'learning_rate': 1.7866325380498416e-06, 'kl': 0.0157, 'entropy': -0.0635, 'ce_loss': 0.0132, 'epoch': 0.71}
24%|██▍ | 50/210 [14:31<45:50, 17.19s/it]
24%|██▍ | 51/210 [14:48<45:47, 17.28s/it]
{'loss': 0.0581, 'grad_norm': 0.4117690324783325, 'learning_rate': 1.7769836147278378e-06, 'kl': 0.0097, 'entropy': -0.0216, 'ce_loss': 0.0213, 'epoch': 0.73}
24%|██▍ | 51/210 [14:48<45:47, 17.28s/it]
25%|██▍ | 52/210 [15:05<45:29, 17.27s/it]
{'loss': 0.0532, 'grad_norm': 0.41745489835739136, 'learning_rate': 1.7671486066220965e-06, 'kl': 0.0076, 'entropy': -0.0254, 'ce_loss': 0.0141, 'epoch': 0.74}
25%|██▍ | 52/210 [15:05<45:29, 17.27s/it]
25%|██▌ | 53/210 [15:22<45:04, 17.23s/it]
{'loss': 0.0593, 'grad_norm': 0.43338218331336975, 'learning_rate': 1.7571298691817175e-06, 'kl': 0.0122, 'entropy': -0.026, 'ce_loss': 0.0095, 'epoch': 0.75}
25%|██▌ | 53/210 [15:22<45:04, 17.23s/it]
26%|██▌ | 54/210 [15:40<44:53, 17.26s/it]
{'loss': 0.0524, 'grad_norm': 0.4206220805644989, 'learning_rate': 1.7469298018583166e-06, 'kl': 0.0143, 'entropy': -0.0388, 'ce_loss': 0.0044, 'epoch': 0.77}
26%|██▌ | 54/210 [15:40<44:53, 17.26s/it]
26%|██▌ | 55/210 [15:57<44:36, 17.26s/it]
{'loss': 0.0499, 'grad_norm': 0.3780882656574249, 'learning_rate': 1.736550847531366e-06, 'kl': 0.0092, 'entropy': -0.0515, 'ce_loss': 0.0165, 'epoch': 0.78}
26%|██▌ | 55/210 [15:57<44:36, 17.26s/it]
27%|██▋ | 56/210 [16:14<44:14, 17.24s/it]
{'loss': 0.0511, 'grad_norm': 0.39666762948036194, 'learning_rate': 1.7259954919231307e-06, 'kl': 0.006, 'entropy': -0.0302, 'ce_loss': 0.0091, 'epoch': 0.8}
27%|██▋ | 56/210 [16:14<44:14, 17.24s/it]
27%|██▋ | 57/210 [16:31<43:48, 17.18s/it]
{'loss': 0.055, 'grad_norm': 0.44600924849510193, 'learning_rate': 1.7152662630033503e-06, 'kl': 0.0075, 'entropy': -0.051, 'ce_loss': 0.0165, 'epoch': 0.81}
27%|██▋ | 57/210 [16:31<43:48, 17.18s/it]
28%|██▊ | 58/210 [16:48<43:22, 17.12s/it]
{'loss': 0.05, 'grad_norm': 0.382903516292572, 'learning_rate': 1.704365730383796e-06, 'kl': 0.0084, 'entropy': -0.0645, 'ce_loss': 0.018, 'epoch': 0.83}
28%|██▊ | 58/210 [16:48<43:22, 17.12s/it]
28%|██▊ | 59/210 [17:05<43:03, 17.11s/it]
{'loss': 0.0572, 'grad_norm': 0.40414997935295105, 'learning_rate': 1.6932965047028618e-06, 'kl': 0.0215, 'entropy': -0.1543, 'ce_loss': 0.0058, 'epoch': 0.84}
28%|██▊ | 59/210 [17:05<43:03, 17.11s/it]
29%|██▊ | 60/210 [17:23<42:53, 17.15s/it]
{'loss': 0.0476, 'grad_norm': 0.35673263669013977, 'learning_rate': 1.682061237000322e-06, 'kl': 0.0053, 'entropy': -0.0742, 'ce_loss': 0.014, 'epoch': 0.85}
29%|██▊ | 60/210 [17:23<42:53, 17.15s/it]
29%|██▉ | 61/210 [17:40<42:35, 17.15s/it]
{'loss': 0.0536, 'grad_norm': 0.4322223365306854, 'learning_rate': 1.6706626180824183e-06, 'kl': 0.0067, 'entropy': -0.0125, 'ce_loss': 0.017, 'epoch': 0.87}
29%|██▉ | 61/210 [17:40<42:35, 17.15s/it]
30%|██▉ | 62/210 [17:57<42:30, 17.23s/it]
{'loss': 0.0615, 'grad_norm': 0.4586760103702545, 'learning_rate': 1.6591033778774228e-06, 'kl': 0.0048, 'entropy': -0.0583, 'ce_loss': 0.0161, 'epoch': 0.88}
30%|██▉ | 62/210 [17:57<42:30, 17.23s/it]
30%|███ | 63/210 [18:14<42:03, 17.17s/it]
{'loss': 0.0597, 'grad_norm': 0.4025050401687622, 'learning_rate': 1.6473862847818276e-06, 'kl': 0.0008, 'entropy': -0.0291, 'ce_loss': 0.0117, 'epoch': 0.9}
30%|███ | 63/210 [18:14<42:03, 17.17s/it]
30%|███ | 64/210 [18:31<41:50, 17.20s/it]
{'loss': 0.0545, 'grad_norm': 0.3930206894874573, 'learning_rate': 1.6355141449973253e-06, 'kl': 0.0096, 'entropy': -0.0408, 'ce_loss': 0.0098, 'epoch': 0.91}
30%|███ | 64/210 [18:31<41:50, 17.20s/it]
31%|███ | 65/210 [18:49<41:30, 17.18s/it]
{'loss': 0.0584, 'grad_norm': 0.43545281887054443, 'learning_rate': 1.6234898018587336e-06, 'kl': 0.0002, 'entropy': -0.0084, 'ce_loss': 0.0182, 'epoch': 0.93}
31%|███ | 65/210 [18:49<41:30, 17.18s/it]
31%|███▏ | 66/210 [19:06<41:09, 17.15s/it]
{'loss': 0.0473, 'grad_norm': 0.3921285569667816, 'learning_rate': 1.6113161351530257e-06, 'kl': 0.0099, 'entropy': -0.0298, 'ce_loss': 0.0098, 'epoch': 0.94}
31%|███▏ | 66/210 [19:06<41:09, 17.15s/it]
32%|███▏ | 67/210 [19:23<41:05, 17.24s/it]
{'loss': 0.056, 'grad_norm': 0.4304642379283905, 'learning_rate': 1.5989960604296337e-06, 'kl': 0.0131, 'entropy': -0.0292, 'ce_loss': 0.0102, 'epoch': 0.95}
32%|███▏ | 67/210 [19:23<41:05, 17.24s/it]
32%|███▏ | 68/210 [19:41<40:54, 17.29s/it]
{'loss': 0.0626, 'grad_norm': 0.4403439462184906, 'learning_rate': 1.5865325283021828e-06, 'kl': 0.007, 'entropy': -0.0933, 'ce_loss': 0.0168, 'epoch': 0.97}
32%|███▏ | 68/210 [19:41<40:54, 17.29s/it]
33%|███▎ | 69/210 [19:57<40:22, 17.18s/it]
{'loss': 0.0537, 'grad_norm': 0.4378717839717865, 'learning_rate': 1.573928523741832e-06, 'kl': 0.0068, 'entropy': -0.0131, 'ce_loss': 0.0166, 'epoch': 0.98}
33%|███▎ | 69/210 [19:57<40:22, 17.18s/it]
33%|███▎ | 70/210 [20:15<39:59, 17.14s/it]
{'loss': 0.0533, 'grad_norm': 0.4272703230381012, 'learning_rate': 1.5611870653623825e-06, 'kl': 0.0054, 'entropy': -0.1128, 'ce_loss': 0.008, 'epoch': 1.0}
33%|███▎ | 70/210 [20:15<39:59, 17.14s/it]
34%|███▍ | 71/210 [20:17<29:21, 12.67s/it]
{'loss': 0.0556, 'grad_norm': 0.4272703230381012, 'learning_rate': 1.5483112046973307e-06, 'kl': 0.0096, 'entropy': -0.0474, 'ce_loss': 0.0461, 'epoch': 1.0}
34%|███▍ | 71/210 [20:17<29:21, 12.67s/it]
34%|███▍ | 72/210 [20:34<32:21, 14.07s/it]
{'loss': 0.0412, 'grad_norm': 0.9038904905319214, 'learning_rate': 1.5353040254690392e-06, 'kl': 0.0109, 'entropy': -0.0513, 'ce_loss': 0.0054, 'epoch': 1.01}
34%|███▍ | 72/210 [20:34<32:21, 14.07s/it]
35%|███▍ | 73/210 [20:51<34:22, 15.06s/it]
{'loss': 0.0461, 'grad_norm': 0.3670438826084137, 'learning_rate': 1.5221686428501928e-06, 'kl': 0.0172, 'entropy': 0.0074, 'ce_loss': 0.018, 'epoch': 1.03}
35%|███▍ | 73/210 [20:51<34:22, 15.06s/it]
35%|███▌ | 74/210 [21:08<35:23, 15.61s/it]
{'loss': 0.0408, 'grad_norm': 0.36899104714393616, 'learning_rate': 1.5089082027177289e-06, 'kl': 0.0157, 'entropy': -0.064, 'ce_loss': 0.0128, 'epoch': 1.04}
35%|███▌ | 74/210 [21:08<35:23, 15.61s/it]
36%|███▌ | 75/210 [21:26<36:20, 16.16s/it]
{'loss': 0.0441, 'grad_norm': 0.3659462630748749, 'learning_rate': 1.4955258808994094e-06, 'kl': 0.021, 'entropy': -0.0242, 'ce_loss': 0.0061, 'epoch': 1.06}
36%|███▌ | 75/210 [21:26<36:20, 16.16s/it]
36%|███▌ | 76/210 [21:43<36:46, 16.47s/it]
{'loss': 0.0436, 'grad_norm': 0.3586432635784149, 'learning_rate': 1.482024882413222e-06, 'kl': 0.0444, 'entropy': -0.0242, 'ce_loss': 0.0146, 'epoch': 1.07}
36%|███▌ | 76/210 [21:43<36:46, 16.47s/it]
37%|███▋ | 77/210 [22:00<36:53, 16.64s/it]
{'loss': 0.0452, 'grad_norm': 0.40676310658454895, 'learning_rate': 1.4684084406997901e-06, 'kl': 0.0283, 'entropy': -0.0703, 'ce_loss': 0.0093, 'epoch': 1.09}
37%|███▋ | 77/210 [22:00<36:53, 16.64s/it]
37%|███▋ | 78/210 [22:17<36:55, 16.79s/it]
{'loss': 0.0409, 'grad_norm': 0.363031268119812, 'learning_rate': 1.4546798168479756e-06, 'kl': 0.0184, 'entropy': -0.0466, 'ce_loss': 0.0115, 'epoch': 1.1}
37%|███▋ | 78/210 [22:17<36:55, 16.79s/it]
38%|███▊ | 79/210 [22:34<36:53, 16.89s/it]
{'loss': 0.0432, 'grad_norm': 0.4143624007701874, 'learning_rate': 1.4408422988138582e-06, 'kl': 0.0135, 'entropy': -0.05, 'ce_loss': 0.0085, 'epoch': 1.11}
38%|███▊ | 79/210 [22:34<36:53, 16.89s/it]
38%|███▊ | 80/210 [22:51<36:47, 16.98s/it]
{'loss': 0.0374, 'grad_norm': 0.344946026802063, 'learning_rate': 1.4268992006332845e-06, 'kl': 0.0221, 'entropy': -0.0747, 'ce_loss': 0.0089, 'epoch': 1.13}
38%|███▊ | 80/210 [22:51<36:47, 16.98s/it]
39%|███▊ | 81/210 [23:09<36:43, 17.08s/it]
{'loss': 0.0442, 'grad_norm': 0.3750915825366974, 'learning_rate': 1.4128538616281659e-06, 'kl': 0.0146, 'entropy': -0.01, 'ce_loss': 0.0042, 'epoch': 1.14}
39%|███▊ | 81/210 [23:09<36:43, 17.08s/it]
39%|███▉ | 82/210 [23:26<36:29, 17.11s/it]
{'loss': 0.0419, 'grad_norm': 0.3898021876811981, 'learning_rate': 1.3987096456067233e-06, 'kl': 0.028, 'entropy': -0.0075, 'ce_loss': 0.0188, 'epoch': 1.16}
39%|███▉ | 82/210 [23:26<36:29, 17.11s/it]
40%|███▉ | 83/210 [23:43<36:20, 17.17s/it]
{'loss': 0.0433, 'grad_norm': 0.42842376232147217, 'learning_rate': 1.3844699400578693e-06, 'kl': 0.0214, 'entropy': -0.1064, 'ce_loss': 0.0038, 'epoch': 1.17}
40%|███▉ | 83/210 [23:43<36:20, 17.17s/it]
40%|████ | 84/210 [24:00<36:02, 17.17s/it]
{'loss': 0.0418, 'grad_norm': 0.40163445472717285, 'learning_rate': 1.3701381553399146e-06, 'kl': 0.0229, 'entropy': -0.0139, 'ce_loss': 0.0056, 'epoch': 1.19}
40%|████ | 84/210 [24:00<36:02, 17.17s/it]
40%|████ | 85/210 [24:17<35:41, 17.13s/it]
{'loss': 0.0433, 'grad_norm': 0.424587607383728, 'learning_rate': 1.3557177238637985e-06, 'kl': 0.0239, 'entropy': -0.0398, 'ce_loss': 0.0067, 'epoch': 1.2}
40%|████ | 85/210 [24:17<35:41, 17.13s/it]
41%|████ | 86/210 [24:35<35:29, 17.17s/it]
{'loss': 0.0435, 'grad_norm': 0.45604848861694336, 'learning_rate': 1.3412120992710422e-06, 'kl': 0.0131, 'entropy': -0.0349, 'ce_loss': 0.0075, 'epoch': 1.21}
41%|████ | 86/210 [24:35<35:29, 17.17s/it]
41%|████▏ | 87/210 [24:52<35:11, 17.16s/it]
{'loss': 0.0399, 'grad_norm': 0.43648621439933777, 'learning_rate': 1.3266247556066122e-06, 'kl': 0.0237, 'entropy': -0.0103, 'ce_loss': 0.0144, 'epoch': 1.23}
41%|████▏ | 87/210 [24:52<35:11, 17.16s/it]
42%|████▏ | 88/210 [25:09<34:51, 17.14s/it]
{'loss': 0.0402, 'grad_norm': 0.4169567823410034, 'learning_rate': 1.3119591864868977e-06, 'kl': 0.0292, 'entropy': -0.0752, 'ce_loss': 0.0067, 'epoch': 1.24}
42%|████▏ | 88/210 [25:09<34:51, 17.14s/it]
42%|████▏ | 89/210 [25:26<34:33, 17.14s/it]
{'loss': 0.0396, 'grad_norm': 0.40840384364128113, 'learning_rate': 1.2972189042630042e-06, 'kl': 0.0171, 'entropy': -0.0762, 'ce_loss': 0.0067, 'epoch': 1.26}
42%|████▏ | 89/210 [25:26<34:33, 17.14s/it]
43%|████▎ | 90/210 [25:43<34:18, 17.15s/it]
{'loss': 0.044, 'grad_norm': 0.41889750957489014, 'learning_rate': 1.282407439179557e-06, 'kl': 0.0228, 'entropy': -0.0349, 'ce_loss': 0.012, 'epoch': 1.27}
43%|████▎ | 90/210 [25:43<34:18, 17.15s/it]
43%|████▎ | 91/210 [26:00<33:59, 17.14s/it]
{'loss': 0.0356, 'grad_norm': 0.41735032200813293, 'learning_rate': 1.2675283385292211e-06, 'kl': 0.015, 'entropy': -0.0859, 'ce_loss': 0.0074, 'epoch': 1.28}
43%|████▎ | 91/210 [26:00<33:59, 17.14s/it]
44%|████▍ | 92/210 [26:18<33:43, 17.15s/it]
{'loss': 0.0395, 'grad_norm': 0.4181468188762665, 'learning_rate': 1.252585165803135e-06, 'kl': 0.0188, 'entropy': -0.0625, 'ce_loss': 0.01, 'epoch': 1.3}
44%|████▍ | 92/210 [26:18<33:43, 17.15s/it]
44%|████▍ | 93/210 [26:35<33:26, 17.15s/it]
{'loss': 0.0364, 'grad_norm': 0.3777095079421997, 'learning_rate': 1.2375814998374711e-06, 'kl': 0.0182, 'entropy': -0.0356, 'ce_loss': 0.0154, 'epoch': 1.31}
44%|████▍ | 93/210 [26:35<33:26, 17.15s/it]
45%|████▍ | 94/210 [26:52<33:09, 17.15s/it]
{'loss': 0.0413, 'grad_norm': 0.41448715329170227, 'learning_rate': 1.2225209339563143e-06, 'kl': 0.0322, 'entropy': -0.1387, 'ce_loss': 0.0134, 'epoch': 1.33}
45%|████▍ | 94/210 [26:52<33:09, 17.15s/it]
45%|████▌ | 95/210 [27:09<32:49, 17.12s/it]
{'loss': 0.0387, 'grad_norm': 0.4828692674636841, 'learning_rate': 1.207407075111075e-06, 'kl': 0.026, 'entropy': -0.1182, 'ce_loss': 0.0069, 'epoch': 1.34}
45%|████▌ | 95/210 [27:09<32:49, 17.12s/it]
46%|████▌ | 96/210 [27:26<32:32, 17.13s/it]
{'loss': 0.0441, 'grad_norm': 0.3837016820907593, 'learning_rate': 1.1922435430166369e-06, 'kl': 0.0171, 'entropy': -0.0942, 'ce_loss': 0.0095, 'epoch': 1.36}
46%|████▌ | 96/210 [27:26<32:32, 17.13s/it]
46%|████▌ | 97/210 [27:44<32:31, 17.27s/it]
{'loss': 0.0382, 'grad_norm': 0.4030384123325348, 'learning_rate': 1.1770339692844483e-06, 'kl': 0.0238, 'entropy': -0.1416, 'ce_loss': 0.0113, 'epoch': 1.37}
46%|████▌ | 97/210 [27:44<32:31, 17.27s/it]
47%|████▋ | 98/210 [28:01<32:13, 17.26s/it]
{'loss': 0.0395, 'grad_norm': 0.3893054127693176, 'learning_rate': 1.161781996552765e-06, 'kl': 0.033, 'entropy': -0.026, 'ce_loss': 0.014, 'epoch': 1.38}
47%|████▋ | 98/210 [28:01<32:13, 17.26s/it]
47%|████▋ | 99/210 [28:18<31:56, 17.26s/it]
{'loss': 0.0421, 'grad_norm': 0.3755781054496765, 'learning_rate': 1.1464912776142492e-06, 'kl': 0.0289, 'entropy': -0.0476, 'ce_loss': 0.0091, 'epoch': 1.4}
47%|████▋ | 99/210 [28:18<31:56, 17.26s/it]
48%|████▊ | 100/210 [28:36<31:40, 17.27s/it]
{'loss': 0.0391, 'grad_norm': 0.36402571201324463, 'learning_rate': 1.1311654745411422e-06, 'kl': 0.0342, 'entropy': -0.0055, 'ce_loss': 0.0072, 'epoch': 1.41}
48%|████▊ | 100/210 [28:36<31:40, 17.27s/it]
48%|████▊ | 101/210 [28:53<31:24, 17.29s/it]
{'loss': 0.0486, 'grad_norm': 0.48946553468704224, 'learning_rate': 1.115808257808209e-06, 'kl': 0.0096, 'entropy': -0.0259, 'ce_loss': 0.0046, 'epoch': 1.43}
48%|████▊ | 101/210 [28:53<31:24, 17.29s/it]
49%|████▊ | 102/210 [29:10<31:02, 17.25s/it]
{'loss': 0.0397, 'grad_norm': 0.3809380829334259, 'learning_rate': 1.1004233054136725e-06, 'kl': 0.008, 'entropy': -0.0574, 'ce_loss': 0.0081, 'epoch': 1.44}
49%|████▊ | 102/210 [29:10<31:02, 17.25s/it]
49%|████▉ | 103/210 [29:27<30:35, 17.16s/it]
{'loss': 0.0462, 'grad_norm': 0.42104533314704895, 'learning_rate': 1.0850143019983472e-06, 'kl': 0.0229, 'entropy': -0.0334, 'ce_loss': 0.0109, 'epoch': 1.46}
49%|████▉ | 103/210 [29:27<30:35, 17.16s/it]
50%|████▉ | 104/210 [29:44<30:23, 17.21s/it]
{'loss': 0.0405, 'grad_norm': 0.3753838837146759, 'learning_rate': 1.0695849379631813e-06, 'kl': 0.0122, 'entropy': -0.0273, 'ce_loss': 0.0109, 'epoch': 1.47}
50%|████▉ | 104/210 [29:44<30:23, 17.21s/it]
50%|█████ | 105/210 [30:02<30:08, 17.22s/it]
{'loss': 0.0379, 'grad_norm': 0.3733779191970825, 'learning_rate': 1.0541389085854176e-06, 'kl': 0.0159, 'entropy': -0.0052, 'ce_loss': 0.0177, 'epoch': 1.48}
50%|█████ | 105/210 [30:02<30:08, 17.22s/it]
50%|█████ | 106/210 [30:19<29:48, 17.20s/it]
{'loss': 0.0354, 'grad_norm': 0.3768526017665863, 'learning_rate': 1.0386799131335887e-06, 'kl': 0.02, 'entropy': -0.0615, 'ce_loss': 0.0073, 'epoch': 1.5}
50%|█████ | 106/210 [30:19<29:48, 17.20s/it]
51%|█████ | 107/210 [30:36<29:34, 17.23s/it]
{'loss': 0.0398, 'grad_norm': 0.3560822010040283, 'learning_rate': 1.0232116539815556e-06, 'kl': 0.0211, 'entropy': -0.0273, 'ce_loss': 0.01, 'epoch': 1.51}
51%|█████ | 107/210 [30:36<29:34, 17.23s/it]
51%|█████▏ | 108/210 [30:53<29:20, 17.26s/it]
{'loss': 0.0371, 'grad_norm': 0.3768574595451355, 'learning_rate': 1.0077378357218021e-06, 'kl': 0.0124, 'entropy': -0.0239, 'ce_loss': 0.0067, 'epoch': 1.53}
51%|█████▏ | 108/210 [30:53<29:20, 17.26s/it]
52%|█████▏ | 109/210 [31:10<29:00, 17.23s/it]
{'loss': 0.0427, 'grad_norm': 0.4112342298030853, 'learning_rate': 9.922621642781978e-07, 'kl': 0.0254, 'entropy': -0.0159, 'ce_loss': 0.0039, 'epoch': 1.54}
52%|█████▏ | 109/210 [31:10<29:00, 17.23s/it]
52%|█████▏ | 110/210 [31:28<28:42, 17.23s/it]
{'loss': 0.0395, 'grad_norm': 0.3958025574684143, 'learning_rate': 9.767883460184441e-07, 'kl': 0.0371, 'entropy': -0.0674, 'ce_loss': 0.0102, 'epoch': 1.56}
52%|█████▏ | 110/210 [31:28<28:42, 17.23s/it]
53%|█████▎ | 111/210 [31:45<28:18, 17.16s/it]
{'loss': 0.0421, 'grad_norm': 0.3947964310646057, 'learning_rate': 9.613200868664112e-07, 'kl': 0.0168, 'entropy': 0.0203, 'ce_loss': 0.0089, 'epoch': 1.57}
53%|█████▎ | 111/210 [31:45<28:18, 17.16s/it]
53%|█████▎ | 112/210 [32:02<28:00, 17.15s/it]
{'loss': 0.0407, 'grad_norm': 0.3924732208251953, 'learning_rate': 9.458610914145825e-07, 'kl': 0.0145, 'entropy': -0.0732, 'ce_loss': 0.0061, 'epoch': 1.58}
53%|█████▎ | 112/210 [32:02<28:00, 17.15s/it]
54%|█████▍ | 113/210 [32:19<27:39, 17.11s/it]
{'loss': 0.0377, 'grad_norm': 0.3410203754901886, 'learning_rate': 9.304150620368187e-07, 'kl': 0.0123, 'entropy': -0.0527, 'ce_loss': 0.0079, 'epoch': 1.6}
54%|█████▍ | 113/210 [32:19<27:39, 17.11s/it]
54%|█████▍ | 114/210 [32:36<27:23, 17.12s/it]
{'loss': 0.0401, 'grad_norm': 0.37413278222084045, 'learning_rate': 9.149856980016528e-07, 'kl': 0.0166, 'entropy': -0.0894, 'ce_loss': 0.011, 'epoch': 1.61}
54%|█████▍ | 114/210 [32:36<27:23, 17.12s/it]
55%|█████▍ | 115/210 [32:53<27:02, 17.08s/it]
{'loss': 0.0427, 'grad_norm': 0.3672293424606323, 'learning_rate': 8.995766945863275e-07, 'kl': 0.025, 'entropy': 0.019, 'ce_loss': 0.0106, 'epoch': 1.63}
55%|█████▍ | 115/210 [32:53<27:02, 17.08s/it]
55%|█████▌ | 116/210 [33:10<26:48, 17.11s/it]
{'loss': 0.0434, 'grad_norm': 0.37829938530921936, 'learning_rate': 8.841917421917912e-07, 'kl': 0.0272, 'entropy': 0.0211, 'ce_loss': 0.0229, 'epoch': 1.64}
55%|█████▌ | 116/210 [33:10<26:48, 17.11s/it]
56%|█████▌ | 117/210 [33:27<26:33, 17.14s/it]
{'loss': 0.0375, 'grad_norm': 0.4009958803653717, 'learning_rate': 8.688345254588577e-07, 'kl': 0.0089, 'entropy': 0.0112, 'ce_loss': 0.0066, 'epoch': 1.65}
56%|█████▌ | 117/210 [33:27<26:33, 17.14s/it]
56%|█████▌ | 118/210 [33:45<26:20, 17.18s/it]
{'loss': 0.0395, 'grad_norm': 0.38676518201828003, 'learning_rate': 8.535087223857508e-07, 'kl': 0.01, 'entropy': 0.0054, 'ce_loss': 0.0064, 'epoch': 1.67}
56%|█████▌ | 118/210 [33:45<26:20, 17.18s/it]
57%|█████▋ | 119/210 [34:02<26:04, 17.20s/it]
{'loss': 0.042, 'grad_norm': 0.383090078830719, 'learning_rate': 8.382180034472353e-07, 'kl': 0.0398, 'entropy': -0.0659, 'ce_loss': 0.03, 'epoch': 1.68}
57%|█████▋ | 119/210 [34:02<26:04, 17.20s/it]
57%|█████▋ | 120/210 [34:19<25:45, 17.17s/it]
{'loss': 0.0358, 'grad_norm': 0.3907102048397064, 'learning_rate': 8.229660307155517e-07, 'kl': 0.0125, 'entropy': -0.0908, 'ce_loss': 0.0064, 'epoch': 1.7}
57%|█████▋ | 120/210 [34:19<25:45, 17.17s/it]
58%|█████▊ | 121/210 [34:36<25:31, 17.20s/it]
{'loss': 0.0401, 'grad_norm': 0.3737153112888336, 'learning_rate': 8.077564569833631e-07, 'kl': 0.0219, 'entropy': -0.0742, 'ce_loss': 0.0063, 'epoch': 1.71}
58%|█████▊ | 121/210 [34:36<25:31, 17.20s/it]
58%|█████▊ | 122/210 [34:53<25:11, 17.18s/it]
{'loss': 0.0434, 'grad_norm': 0.44650998711586, 'learning_rate': 7.92592924888925e-07, 'kl': 0.0208, 'entropy': -0.0525, 'ce_loss': 0.0106, 'epoch': 1.73}
58%|█████▊ | 122/210 [34:53<25:11, 17.18s/it]
59%|█████▊ | 123/210 [35:10<24:53, 17.16s/it]
{'loss': 0.0405, 'grad_norm': 0.4064434766769409, 'learning_rate': 7.774790660436857e-07, 'kl': 0.0177, 'entropy': -0.0166, 'ce_loss': 0.0047, 'epoch': 1.74}
59%|█████▊ | 123/210 [35:10<24:53, 17.16s/it]
59%|█████▉ | 124/210 [35:28<24:35, 17.16s/it]
{'loss': 0.0492, 'grad_norm': 0.46531906723976135, 'learning_rate': 7.624185001625291e-07, 'kl': 0.0121, 'entropy': -0.0244, 'ce_loss': 0.0151, 'epoch': 1.75}
59%|█████▉ | 124/210 [35:28<24:35, 17.16s/it]
60%|█████▉ | 125/210 [35:45<24:14, 17.11s/it]
{'loss': 0.0443, 'grad_norm': 0.47013038396835327, 'learning_rate': 7.474148341968651e-07, 'kl': 0.0117, 'entropy': -0.063, 'ce_loss': 0.0102, 'epoch': 1.77}
60%|█████▉ | 125/210 [35:45<24:14, 17.11s/it]
60%|██████ | 126/210 [36:02<23:56, 17.11s/it]
{'loss': 0.0376, 'grad_norm': 0.411589115858078, 'learning_rate': 7.324716614707793e-07, 'kl': 0.0092, 'entropy': -0.0262, 'ce_loss': 0.0055, 'epoch': 1.78}
60%|██████ | 126/210 [36:02<23:56, 17.11s/it]
60%|██████ | 127/210 [36:19<23:50, 17.24s/it]
{'loss': 0.0385, 'grad_norm': 0.35508039593696594, 'learning_rate': 7.175925608204427e-07, 'kl': 0.0227, 'entropy': -0.0317, 'ce_loss': 0.0069, 'epoch': 1.8}
60%|██████ | 127/210 [36:19<23:50, 17.24s/it]
61%|██████ | 128/210 [36:37<23:35, 17.26s/it]
{'loss': 0.0416, 'grad_norm': 0.38547420501708984, 'learning_rate': 7.027810957369956e-07, 'kl': 0.0145, 'entropy': -0.0483, 'ce_loss': 0.0175, 'epoch': 1.81}
61%|██████ | 128/210 [36:37<23:35, 17.26s/it]
61%|██████▏ | 129/210 [36:54<23:14, 17.22s/it]
{'loss': 0.042, 'grad_norm': 0.38511332869529724, 'learning_rate': 6.880408135131022e-07, 'kl': 0.0177, 'entropy': 0.0044, 'ce_loss': 0.018, 'epoch': 1.83}
61%|██████▏ | 129/210 [36:54<23:14, 17.22s/it]
62%|██████▏ | 130/210 [37:11<23:00, 17.25s/it]
{'loss': 0.0419, 'grad_norm': 0.39483943581581116, 'learning_rate': 6.733752443933878e-07, 'kl': 0.0143, 'entropy': -0.051, 'ce_loss': 0.0098, 'epoch': 1.84}
62%|██████▏ | 130/210 [37:11<23:00, 17.25s/it]
62%|██████▏ | 131/210 [37:28<22:43, 17.26s/it]
{'loss': 0.0382, 'grad_norm': 0.37223953008651733, 'learning_rate': 6.587879007289575e-07, 'kl': 0.0591, 'entropy': -0.0591, 'ce_loss': 0.0102, 'epoch': 1.85}
62%|██████▏ | 131/210 [37:28<22:43, 17.26s/it]
63%|██████▎ | 132/210 [37:45<22:21, 17.20s/it]
{'loss': 0.0502, 'grad_norm': 0.43174198269844055, 'learning_rate': 6.442822761362014e-07, 'kl': 0.1226, 'entropy': 0.0109, 'ce_loss': 0.0242, 'epoch': 1.87}
63%|██████▎ | 132/210 [37:45<22:21, 17.20s/it]
63%|██████▎ | 133/210 [38:03<22:03, 17.19s/it]
{'loss': 0.0393, 'grad_norm': 0.38623350858688354, 'learning_rate': 6.298618446600855e-07, 'kl': 0.0238, 'entropy': -0.127, 'ce_loss': 0.0121, 'epoch': 1.88}
63%|██████▎ | 133/210 [38:03<22:03, 17.19s/it]
64%|██████▍ | 134/210 [38:20<21:45, 17.18s/it]
{'loss': 0.0393, 'grad_norm': 0.40570443868637085, 'learning_rate': 6.155300599421305e-07, 'kl': 0.027, 'entropy': -0.0574, 'ce_loss': 0.0104, 'epoch': 1.9}
64%|██████▍ | 134/210 [38:20<21:45, 17.18s/it]
64%|██████▍ | 135/210 [38:37<21:31, 17.22s/it]
{'loss': 0.0431, 'grad_norm': 0.41379815340042114, 'learning_rate': 6.012903543932766e-07, 'kl': 0.0508, 'entropy': -0.0708, 'ce_loss': 0.0174, 'epoch': 1.91}
64%|██████▍ | 135/210 [38:37<21:31, 17.22s/it]
65%|██████▍ | 136/210 [38:54<21:12, 17.20s/it]
{'loss': 0.0441, 'grad_norm': 0.41139155626296997, 'learning_rate': 5.871461383718344e-07, 'kl': 0.0161, 'entropy': -0.0312, 'ce_loss': 0.0152, 'epoch': 1.93}
65%|██████▍ | 136/210 [38:54<21:12, 17.20s/it]
65%|██████▌ | 137/210 [39:11<20:53, 17.17s/it]
{'loss': 0.0415, 'grad_norm': 0.4377204477787018, 'learning_rate': 5.731007993667154e-07, 'kl': 0.0459, 'entropy': -0.0796, 'ce_loss': 0.0062, 'epoch': 1.94}
65%|██████▌ | 137/210 [39:11<20:53, 17.17s/it]
66%|██████▌ | 138/210 [39:28<20:33, 17.13s/it]
{'loss': 0.0406, 'grad_norm': 0.3920570909976959, 'learning_rate': 5.591577011861419e-07, 'kl': 0.0253, 'entropy': -0.0457, 'ce_loss': 0.0109, 'epoch': 1.95}
66%|██████▌ | 138/210 [39:28<20:33, 17.13s/it]
66%|██████▌ | 139/210 [39:45<20:16, 17.13s/it]
{'loss': 0.042, 'grad_norm': 0.44121894240379333, 'learning_rate': 5.453201831520245e-07, 'kl': 0.0532, 'entropy': -0.0781, 'ce_loss': 0.0114, 'epoch': 1.97}
66%|██████▌ | 139/210 [39:45<20:16, 17.13s/it]
67%|██████▋ | 140/210 [40:03<20:01, 17.16s/it]
{'loss': 0.0356, 'grad_norm': 0.37643083930015564, 'learning_rate': 5.3159155930021e-07, 'kl': 0.0152, 'entropy': -0.0557, 'ce_loss': 0.005, 'epoch': 1.98}
67%|██████▋ | 140/210 [40:03<20:01, 17.16s/it]
67%|██████▋ | 141/210 [40:20<19:46, 17.19s/it]
{'loss': 0.044, 'grad_norm': 0.416394978761673, 'learning_rate': 5.179751175867784e-07, 'kl': 0.052, 'entropy': 0.0035, 'ce_loss': 0.0073, 'epoch': 2.0}
67%|██████▋ | 141/210 [40:20<19:46, 17.19s/it]
68%|██████▊ | 142/210 [40:22<14:26, 12.75s/it]
{'loss': 0.0334, 'grad_norm': 0.416394978761673, 'learning_rate': 5.044741191005908e-07, 'kl': 0.0129, 'entropy': -0.082, 'ce_loss': 0.0376, 'epoch': 2.0}
68%|██████▊ | 142/210 [40:22<14:26, 12.75s/it]
68%|██████▊ | 143/210 [40:40<15:46, 14.12s/it]
{'loss': 0.0357, 'grad_norm': 0.7566497325897217, 'learning_rate': 4.910917972822712e-07, 'kl': 0.0262, 'entropy': -0.028, 'ce_loss': 0.0096, 'epoch': 2.01}
68%|██████▊ | 143/210 [40:40<15:46, 14.12s/it]
69%|██████▊ | 144/210 [40:57<16:30, 15.01s/it]
{'loss': 0.036, 'grad_norm': 0.3802626132965088, 'learning_rate': 4.778313571498073e-07, 'kl': 0.0276, 'entropy': -0.0206, 'ce_loss': 0.009, 'epoch': 2.03}
69%|██████▊ | 144/210 [40:57<16:30, 15.01s/it]
69%|██████▉ | 145/210 [41:14<16:58, 15.66s/it]
{'loss': 0.0326, 'grad_norm': 0.3390200734138489, 'learning_rate': 4.646959745309609e-07, 'kl': 0.0211, 'entropy': 0.0483, 'ce_loss': 0.0086, 'epoch': 2.04}
69%|██████▉ | 145/210 [41:14<16:58, 15.66s/it]
70%|██████▉ | 146/210 [41:31<17:10, 16.10s/it]
{'loss': 0.0299, 'grad_norm': 0.33013781905174255, 'learning_rate': 4.51688795302669e-07, 'kl': 0.0215, 'entropy': -0.0562, 'ce_loss': 0.0039, 'epoch': 2.06}
70%|██████▉ | 146/210 [41:31<17:10, 16.10s/it]
70%|███████ | 147/210 [41:48<17:13, 16.40s/it]
{'loss': 0.0334, 'grad_norm': 0.3755772113800049, 'learning_rate': 4.388129346376177e-07, 'kl': 0.0466, 'entropy': -0.0217, 'ce_loss': 0.0088, 'epoch': 2.07}
70%|███████ | 147/210 [41:48<17:13, 16.40s/it]
70%|███████ | 148/210 [42:05<17:09, 16.61s/it]
{'loss': 0.0313, 'grad_norm': 0.32724547386169434, 'learning_rate': 4.2607147625816766e-07, 'kl': 0.0195, 'entropy': -0.0757, 'ce_loss': 0.0053, 'epoch': 2.09}
70%|███████ | 148/210 [42:05<17:09, 16.61s/it]
71%|███████ | 149/210 [42:23<17:09, 16.87s/it]
{'loss': 0.0371, 'grad_norm': 0.3594580292701721, 'learning_rate': 4.134674716978169e-07, 'kl': 0.032, 'entropy': -0.0262, 'ce_loss': 0.009, 'epoch': 2.1}
71%|███████ | 149/210 [42:23<17:09, 16.87s/it]
71%|███████▏ | 150/210 [42:40<17:02, 17.03s/it]
{'loss': 0.0329, 'grad_norm': 0.3714512586593628, 'learning_rate': 4.0100393957036636e-07, 'kl': 0.0146, 'entropy': 0.0142, 'ce_loss': 0.0095, 'epoch': 2.11}
71%|███████▏ | 150/210 [42:40<17:02, 17.03s/it]
72%|███████▏ | 151/210 [42:58<16:52, 17.15s/it]
{'loss': 0.0275, 'grad_norm': 0.3298415541648865, 'learning_rate': 3.8868386484697413e-07, 'kl': 0.0605, 'entropy': -0.1533, 'ce_loss': 0.0235, 'epoch': 2.13}
72%|███████▏ | 151/210 [42:58<16:52, 17.15s/it]
72%|███████▏ | 152/210 [43:15<16:35, 17.17s/it]
{'loss': 0.0295, 'grad_norm': 0.3266574442386627, 'learning_rate': 3.765101981412665e-07, 'kl': 0.0159, 'entropy': -0.1025, 'ce_loss': 0.0041, 'epoch': 2.14}
72%|███████▏ | 152/210 [43:15<16:35, 17.17s/it]
73%|███████▎ | 153/210 [43:32<16:18, 17.16s/it]
{'loss': 0.0313, 'grad_norm': 0.33786967396736145, 'learning_rate': 3.644858550026748e-07, 'kl': 0.0179, 'entropy': -0.0283, 'ce_loss': 0.0094, 'epoch': 2.16}
73%|███████▎ | 153/210 [43:32<16:18, 17.16s/it]
73%|███████▎ | 154/210 [43:49<16:00, 17.16s/it]
{'loss': 0.0297, 'grad_norm': 0.37679359316825867, 'learning_rate': 3.526137152181724e-07, 'kl': 0.0479, 'entropy': -0.0615, 'ce_loss': 0.0079, 'epoch': 2.17}
73%|███████▎ | 154/210 [43:49<16:00, 17.16s/it]
74%|███████▍ | 155/210 [44:06<15:44, 17.17s/it]
{'loss': 0.0355, 'grad_norm': 0.3669978082180023, 'learning_rate': 3.408966221225773e-07, 'kl': 0.0204, 'entropy': -0.0483, 'ce_loss': 0.0051, 'epoch': 2.19}
74%|███████▍ | 155/210 [44:06<15:44, 17.17s/it]
74%|███████▍ | 156/210 [44:23<15:26, 17.16s/it]
{'loss': 0.0322, 'grad_norm': 0.3423168361186981, 'learning_rate': 3.293373819175815e-07, 'kl': 0.0452, 'entropy': -0.0342, 'ce_loss': 0.0049, 'epoch': 2.2}
74%|███████▍ | 156/210 [44:23<15:26, 17.16s/it]
75%|███████▍ | 157/210 [44:41<15:12, 17.21s/it]
{'loss': 0.0337, 'grad_norm': 0.32489022612571716, 'learning_rate': 3.1793876299967814e-07, 'kl': 0.0315, 'entropy': 0.0177, 'ce_loss': 0.0101, 'epoch': 2.21}
75%|███████▍ | 157/210 [44:41<15:12, 17.21s/it]
75%|███████▌ | 158/210 [44:58<14:56, 17.24s/it]
{'loss': 0.0341, 'grad_norm': 0.3478025197982788, 'learning_rate': 3.0670349529713814e-07, 'kl': 0.0713, 'entropy': -0.0549, 'ce_loss': 0.0097, 'epoch': 2.23}
75%|███████▌ | 158/210 [44:58<14:56, 17.24s/it]
76%|███████▌ | 159/210 [45:16<14:43, 17.31s/it]
{'loss': 0.0316, 'grad_norm': 0.37568560242652893, 'learning_rate': 2.9563426961620364e-07, 'kl': 0.0229, 'entropy': -0.1196, 'ce_loss': 0.0042, 'epoch': 2.24}
76%|███████▌ | 159/210 [45:16<14:43, 17.31s/it]
76%|███████▌ | 160/210 [45:33<14:24, 17.30s/it]
{'loss': 0.0395, 'grad_norm': 0.3859495222568512, 'learning_rate': 2.847337369966499e-07, 'kl': 0.0364, 'entropy': -0.05, 'ce_loss': 0.0166, 'epoch': 2.26}
76%|███████▌ | 160/210 [45:33<14:24, 17.30s/it]
77%|███████▋ | 161/210 [45:50<14:05, 17.26s/it]
{'loss': 0.0334, 'grad_norm': 0.3939160704612732, 'learning_rate': 2.740045080768694e-07, 'kl': 0.0427, 'entropy': -0.0884, 'ce_loss': 0.0074, 'epoch': 2.27}
77%|███████▋ | 161/210 [45:50<14:05, 17.26s/it]
77%|███████▋ | 162/210 [46:07<13:46, 17.22s/it]
{'loss': 0.0294, 'grad_norm': 0.39411482214927673, 'learning_rate': 2.634491524686341e-07, 'kl': 0.0171, 'entropy': -0.017, 'ce_loss': 0.0056, 'epoch': 2.28}
77%|███████▋ | 162/210 [46:07<13:46, 17.22s/it]
78%|███████▊ | 163/210 [46:24<13:27, 17.19s/it]
{'loss': 0.0361, 'grad_norm': 0.38195353746414185, 'learning_rate': 2.530701981416834e-07, 'kl': 0.0184, 'entropy': -0.0378, 'ce_loss': 0.0087, 'epoch': 2.3}
78%|███████▊ | 163/210 [46:24<13:27, 17.19s/it]
78%|███████▊ | 164/210 [46:41<13:08, 17.14s/it]
{'loss': 0.0368, 'grad_norm': 0.41065844893455505, 'learning_rate': 2.428701308182826e-07, 'kl': 0.0317, 'entropy': 0.03, 'ce_loss': 0.0099, 'epoch': 2.31}
78%|███████▊ | 164/210 [46:41<13:08, 17.14s/it]
79%|███████▊ | 165/210 [46:58<12:50, 17.12s/it]
{'loss': 0.0355, 'grad_norm': 0.3637394309043884, 'learning_rate': 2.3285139337790337e-07, 'kl': 0.0165, 'entropy': -0.0276, 'ce_loss': 0.0088, 'epoch': 2.33}
79%|███████▊ | 165/210 [46:58<12:50, 17.12s/it]
79%|███████▉ | 166/210 [47:16<12:35, 17.18s/it]
{'loss': 0.0327, 'grad_norm': 0.36824455857276917, 'learning_rate': 2.2301638527216194e-07, 'kl': 0.0256, 'entropy': -0.021, 'ce_loss': 0.0101, 'epoch': 2.34}
79%|███████▉ | 166/210 [47:16<12:35, 17.18s/it]
80%|███████▉ | 167/210 [47:33<12:25, 17.33s/it]
{'loss': 0.0342, 'grad_norm': 0.3927820026874542, 'learning_rate': 2.1336746195015843e-07, 'kl': 0.0654, 'entropy': -0.0559, 'ce_loss': 0.0096, 'epoch': 2.36}
80%|███████▉ | 167/210 [47:33<12:25, 17.33s/it]
80%|████████ | 168/210 [47:50<12:06, 17.30s/it]
{'loss': 0.0331, 'grad_norm': 0.37400922179222107, 'learning_rate': 2.0390693429435623e-07, 'kl': 0.0352, 'entropy': -0.001, 'ce_loss': 0.0079, 'epoch': 2.37}
80%|████████ | 168/210 [47:50<12:06, 17.30s/it]
80%|████████ | 169/210 [48:08<11:48, 17.27s/it]
{'loss': 0.0309, 'grad_norm': 0.3711462616920471, 'learning_rate': 1.946370680671341e-07, 'kl': 0.0349, 'entropy': -0.0048, 'ce_loss': 0.0099, 'epoch': 2.38}
80%|████████ | 169/210 [48:08<11:48, 17.27s/it]
81%|████████ | 170/210 [48:25<11:28, 17.22s/it]
{'loss': 0.0405, 'grad_norm': 0.4175037443637848, 'learning_rate': 1.85560083368143e-07, 'kl': 0.0203, 'entropy': -0.0261, 'ce_loss': 0.0102, 'epoch': 2.4}
81%|████████ | 170/210 [48:25<11:28, 17.22s/it]
81%|████████▏ | 171/210 [48:42<11:10, 17.18s/it]
{'loss': 0.0339, 'grad_norm': 0.40324172377586365, 'learning_rate': 1.7667815410260178e-07, 'kl': 0.021, 'entropy': -0.0493, 'ce_loss': 0.0167, 'epoch': 2.41}
81%|████████▏ | 171/210 [48:42<11:10, 17.18s/it]
82%|████████▏ | 172/210 [48:59<10:52, 17.16s/it]
{'loss': 0.0279, 'grad_norm': 0.36605483293533325, 'learning_rate': 1.6799340746065326e-07, 'kl': 0.017, 'entropy': -0.025, 'ce_loss': 0.0043, 'epoch': 2.43}
82%|████████▏ | 172/210 [48:59<10:52, 17.16s/it]
82%|████████▏ | 173/210 [49:16<10:34, 17.16s/it]
{'loss': 0.0338, 'grad_norm': 0.3507750332355499, 'learning_rate': 1.595079234079104e-07, 'kl': 0.0303, 'entropy': -0.0547, 'ce_loss': 0.0082, 'epoch': 2.44}
82%|████████▏ | 173/210 [49:16<10:34, 17.16s/it]
83%|████████▎ | 174/210 [49:33<10:18, 17.19s/it]
{'loss': 0.029, 'grad_norm': 0.36578890681266785, 'learning_rate': 1.5122373418731304e-07, 'kl': 0.0149, 'entropy': -0.0216, 'ce_loss': 0.0051, 'epoch': 2.46}
83%|████████▎ | 174/210 [49:33<10:18, 17.19s/it]
83%|████████▎ | 175/210 [49:51<10:02, 17.23s/it]
{'loss': 0.0279, 'grad_norm': 0.3522734045982361, 'learning_rate': 1.4314282383241095e-07, 'kl': 0.0593, 'entropy': -0.0957, 'ce_loss': 0.0095, 'epoch': 2.47}
83%|████████▎ | 175/210 [49:51<10:02, 17.23s/it]
84%|████████▍ | 176/210 [50:08<09:47, 17.29s/it]
{'loss': 0.0377, 'grad_norm': 0.3678034543991089, 'learning_rate': 1.3526712769219617e-07, 'kl': 0.0198, 'entropy': -0.0304, 'ce_loss': 0.0044, 'epoch': 2.48}
84%|████████▍ | 176/210 [50:08<09:47, 17.29s/it]
84%|████████▍ | 177/210 [50:26<09:33, 17.37s/it]
{'loss': 0.0375, 'grad_norm': 0.3935472369194031, 'learning_rate': 1.275985319675945e-07, 'kl': 0.0486, 'entropy': -0.0603, 'ce_loss': 0.0082, 'epoch': 2.5}
84%|████████▍ | 177/210 [50:26<09:33, 17.37s/it]
85%|████████▍ | 178/210 [50:43<09:14, 17.31s/it]
{'loss': 0.0338, 'grad_norm': 0.411655992269516, 'learning_rate': 1.201388732597255e-07, 'kl': 0.0277, 'entropy': -0.0732, 'ce_loss': 0.0077, 'epoch': 2.51}
85%|████████▍ | 178/210 [50:43<09:14, 17.31s/it]
85%|████████▌ | 179/210 [51:00<08:55, 17.27s/it]
{'loss': 0.0302, 'grad_norm': 0.3559769093990326, 'learning_rate': 1.1288993813004466e-07, 'kl': 0.015, 'entropy': -0.0344, 'ce_loss': 0.0025, 'epoch': 2.53}
85%|████████▌ | 179/210 [51:00<08:55, 17.27s/it]
86%|████████▌ | 180/210 [51:17<08:37, 17.25s/it]
{'loss': 0.0381, 'grad_norm': 0.3784918785095215, 'learning_rate': 1.0585346267246742e-07, 'kl': 0.0248, 'entropy': -0.0039, 'ce_loss': 0.009, 'epoch': 2.54}
86%|████████▌ | 180/210 [51:17<08:37, 17.25s/it]
86%|████████▌ | 181/210 [51:34<08:19, 17.22s/it]
{'loss': 0.0342, 'grad_norm': 0.4126152992248535, 'learning_rate': 9.903113209758096e-08, 'kl': 0.0439, 'entropy': -0.043, 'ce_loss': 0.0116, 'epoch': 2.56}
86%|████████▌ | 181/210 [51:34<08:19, 17.22s/it]
87%|████████▋ | 182/210 [51:52<08:02, 17.22s/it]
{'loss': 0.0333, 'grad_norm': 0.37268808484077454, 'learning_rate': 9.24245803290431e-08, 'kl': 0.0342, 'entropy': -0.0786, 'ce_loss': 0.0033, 'epoch': 2.57}
87%|████████▋ | 182/210 [51:52<08:02, 17.22s/it]
87%|████████▋ | 183/210 [52:09<07:43, 17.18s/it]
{'loss': 0.0353, 'grad_norm': 0.40694659948349, 'learning_rate': 8.60353896122623e-08, 'kl': 0.0116, 'entropy': -0.0221, 'ce_loss': 0.0049, 'epoch': 2.58}
87%|████████▋ | 183/210 [52:09<07:43, 17.18s/it]
88%|████████▊ | 184/210 [52:26<07:26, 17.18s/it]
{'loss': 0.0388, 'grad_norm': 0.40319371223449707, 'learning_rate': 7.986509013545673e-08, 'kl': 0.0165, 'entropy': -0.0583, 'ce_loss': 0.0054, 'epoch': 2.6}
88%|████████▊ | 184/210 [52:26<07:26, 17.18s/it]
88%|████████▊ | 185/210 [52:43<07:10, 17.23s/it]
{'loss': 0.0289, 'grad_norm': 0.3612806797027588, 'learning_rate': 7.391515966317908e-08, 'kl': 0.0104, 'entropy': -0.0212, 'ce_loss': 0.0049, 'epoch': 2.61}
88%|████████▊ | 185/210 [52:43<07:10, 17.23s/it]
89%|████████▊ | 186/210 [53:01<06:54, 17.26s/it]
{'loss': 0.0331, 'grad_norm': 0.37619268894195557, 'learning_rate': 6.818702318239689e-08, 'kl': 0.0214, 'entropy': -0.104, 'ce_loss': 0.0102, 'epoch': 2.63}
89%|████████▊ | 186/210 [53:01<06:54, 17.26s/it]
89%|████████▉ | 187/210 [53:18<06:36, 17.23s/it]
{'loss': 0.0291, 'grad_norm': 0.34594687819480896, 'learning_rate': 6.268205256121394e-08, 'kl': 0.0349, 'entropy': -0.0515, 'ce_loss': 0.0087, 'epoch': 2.64}
89%|████████▉ | 187/210 [53:18<06:36, 17.23s/it]
90%|████████▉ | 188/210 [53:35<06:19, 17.24s/it]
{'loss': 0.0348, 'grad_norm': 0.36603015661239624, 'learning_rate': 5.7401566220313006e-08, 'kl': 0.0215, 'entropy': -0.0239, 'ce_loss': 0.0048, 'epoch': 2.65}
90%|████████▉ | 188/210 [53:35<06:19, 17.24s/it]
90%|█████████ | 189/210 [53:52<06:01, 17.24s/it]
{'loss': 0.0353, 'grad_norm': 0.3877117931842804, 'learning_rate': 5.234682881719765e-08, 'kl': 0.0679, 'entropy': -0.0723, 'ce_loss': 0.0151, 'epoch': 2.67}
90%|█████████ | 189/210 [53:52<06:01, 17.24s/it]
90%|█████████ | 190/210 [54:09<05:44, 17.21s/it]
{'loss': 0.0306, 'grad_norm': 0.3699752986431122, 'learning_rate': 4.751905094331232e-08, 'kl': 0.0199, 'entropy': -0.0603, 'ce_loss': 0.0138, 'epoch': 2.68}
90%|█████████ | 190/210 [54:09<05:44, 17.21s/it]
91%|█████████ | 191/210 [54:27<05:26, 17.20s/it]
{'loss': 0.0264, 'grad_norm': 0.3301219344139099, 'learning_rate': 4.291938883411006e-08, 'kl': 0.0134, 'entropy': -0.0175, 'ce_loss': 0.0072, 'epoch': 2.7}
91%|█████████ | 191/210 [54:27<05:26, 17.20s/it]
91%|█████████▏| 192/210 [54:44<05:09, 17.17s/it]
{'loss': 0.0301, 'grad_norm': 0.3828076124191284, 'learning_rate': 3.85489440921376e-08, 'kl': 0.0284, 'entropy': -0.0613, 'ce_loss': 0.004, 'epoch': 2.71}
91%|█████████▏| 192/210 [54:44<05:09, 17.17s/it]
92%|█████████▏| 193/210 [55:01<04:53, 17.26s/it]
{'loss': 0.0338, 'grad_norm': 0.35951289534568787, 'learning_rate': 3.440876342320609e-08, 'kl': 0.0354, 'entropy': -0.0383, 'ce_loss': 0.0088, 'epoch': 2.73}
92%|█████████▏| 193/210 [55:01<04:53, 17.26s/it]
92%|█████████▏| 194/210 [55:18<04:36, 17.28s/it]
{'loss': 0.0303, 'grad_norm': 0.37115609645843506, 'learning_rate': 3.049983838570858e-08, 'kl': 0.0359, 'entropy': -0.0869, 'ce_loss': 0.0107, 'epoch': 2.74}
92%|█████████▏| 194/210 [55:18<04:36, 17.28s/it]
93%|█████████▎| 195/210 [55:36<04:20, 17.37s/it]
{'loss': 0.0341, 'grad_norm': 0.3762635588645935, 'learning_rate': 2.682310515314512e-08, 'kl': 0.0383, 'entropy': -0.1094, 'ce_loss': 0.0054, 'epoch': 2.75}
93%|█████████▎| 195/210 [55:36<04:20, 17.37s/it]
93%|█████████▎| 196/210 [55:53<04:03, 17.36s/it]
{'loss': 0.0314, 'grad_norm': 0.4083958864212036, 'learning_rate': 2.337944428991334e-08, 'kl': 0.124, 'entropy': -0.0972, 'ce_loss': 0.0113, 'epoch': 2.77}
93%|█████████▎| 196/210 [55:53<04:03, 17.36s/it]
94%|█████████▍| 197/210 [56:11<03:44, 17.30s/it]
{'loss': 0.0309, 'grad_norm': 0.39548295736312866, 'learning_rate': 2.0169680540415456e-08, 'kl': 0.0134, 'entropy': -0.0076, 'ce_loss': 0.0053, 'epoch': 2.78}
94%|█████████▍| 197/210 [56:11<03:44, 17.30s/it]
94%|█████████▍| 198/210 [56:28<03:26, 17.21s/it]
{'loss': 0.0319, 'grad_norm': 0.3754424750804901, 'learning_rate': 1.7194582631535614e-08, 'kl': 0.085, 'entropy': -0.1396, 'ce_loss': 0.0156, 'epoch': 2.8}
94%|█████████▍| 198/210 [56:28<03:26, 17.21s/it]
95%|█████████▍| 199/210 [56:45<03:09, 17.21s/it]
{'loss': 0.0354, 'grad_norm': 0.3921602666378021, 'learning_rate': 1.4454863088532387e-08, 'kl': 0.0598, 'entropy': -0.0957, 'ce_loss': 0.0137, 'epoch': 2.81}
95%|█████████▍| 199/210 [56:45<03:09, 17.21s/it]
95%|█████████▌| 200/210 [57:02<02:51, 17.18s/it]
{'loss': 0.0312, 'grad_norm': 0.3813055455684662, 'learning_rate': 1.1951178064391499e-08, 'kl': 0.0167, 'entropy': 0.0151, 'ce_loss': 0.0066, 'epoch': 2.83}
95%|█████████▌| 200/210 [57:02<02:51, 17.18s/it]
96%|█████████▌| 201/210 [57:19<02:34, 17.20s/it]
{'loss': 0.035, 'grad_norm': 0.38693967461586, 'learning_rate': 9.684127182679525e-09, 'kl': 0.03, 'entropy': -0.0605, 'ce_loss': 0.0082, 'epoch': 2.84}
96%|█████████▌| 201/210 [57:19<02:34, 17.20s/it]
96%|█████████▌| 202/210 [57:36<02:17, 17.21s/it]
{'loss': 0.0295, 'grad_norm': 0.3731502294540405, 'learning_rate': 7.654253393936439e-09, 'kl': 0.0264, 'entropy': -0.053, 'ce_loss': 0.0114, 'epoch': 2.85}
96%|█████████▌| 202/210 [57:36<02:17, 17.21s/it]
97%|█████████▋| 203/210 [57:54<02:00, 17.24s/it]
{'loss': 0.0328, 'grad_norm': 0.38439086079597473, 'learning_rate': 5.862042845640403e-09, 'kl': 0.0295, 'entropy': -0.0674, 'ce_loss': 0.0084, 'epoch': 2.87}
97%|█████████▋| 203/210 [57:54<02:00, 17.24s/it]
97%|█████████▋| 204/210 [58:11<01:43, 17.25s/it]
{'loss': 0.034, 'grad_norm': 0.3721282482147217, 'learning_rate': 4.3079247657776815e-09, 'kl': 0.0471, 'entropy': -0.0503, 'ce_loss': 0.0069, 'epoch': 2.88}
97%|█████████▋| 204/210 [58:11<01:43, 17.25s/it]
98%|█████████▊| 205/210 [58:28<01:25, 17.18s/it]
{'loss': 0.0314, 'grad_norm': 0.4035966694355011, 'learning_rate': 2.9922713600439852e-09, 'kl': 0.0227, 'entropy': -0.0579, 'ce_loss': 0.0063, 'epoch': 2.9}
98%|█████████▊| 205/210 [58:28<01:25, 17.18s/it]
98%|█████████▊| 206/210 [58:45<01:08, 17.14s/it]
{'loss': 0.0297, 'grad_norm': 0.40415969491004944, 'learning_rate': 1.9153977227022167e-09, 'kl': 0.0439, 'entropy': -0.0522, 'ce_loss': 0.0064, 'epoch': 2.91}
98%|█████████▊| 206/210 [58:45<01:08, 17.14s/it]
99%|█████████▊| 207/210 [59:02<00:51, 17.15s/it]
{'loss': 0.0315, 'grad_norm': 0.3833099603652954, 'learning_rate': 1.0775617611189503e-09, 'kl': 0.0327, 'entropy': -0.0757, 'ce_loss': 0.0118, 'epoch': 2.93}
99%|█████████▊| 207/210 [59:02<00:51, 17.15s/it]
99%|█████████▉| 208/210 [59:19<00:34, 17.19s/it]
{'loss': 0.029, 'grad_norm': 0.3508811593055725, 'learning_rate': 4.789641339963957e-10, 'kl': 0.0225, 'entropy': -0.015, 'ce_loss': 0.0058, 'epoch': 2.94}
99%|█████████▉| 208/210 [59:19<00:34, 17.19s/it]
100%|█████████▉| 209/210 [59:37<00:17, 17.22s/it]
{'loss': 0.0337, 'grad_norm': 0.3555991053581238, 'learning_rate': 1.197482033151731e-10, 'kl': 0.0265, 'entropy': -0.0266, 'ce_loss': 0.0048, 'epoch': 2.95}
100%|█████████▉| 209/210 [59:37<00:17, 17.22s/it]
100%|██████████| 210/210 [59:54<00:00, 17.24s/it]
{'loss': 0.0318, 'grad_norm': 0.4325442314147949, 'learning_rate': 0.0, 'kl': 0.0201, 'entropy': -0.0386, 'ce_loss': 0.0043, 'epoch': 2.97}
100%|██████████| 210/210 [59:54<00:00, 17.24s/it][INFO|trainer.py:2665] 2025-04-16 16:52:15,272 >>
Training completed. Do not forget to share your model on huggingface.co/models =)
{'train_runtime': 3594.5446, 'train_samples_per_second': 1.874, 'train_steps_per_second': 0.058, 'train_loss': 0.04306478500366211, 'epoch': 2.97}
100%|██████████| 210/210 [59:54<00:00, 17.24s/it]
100%|██████████| 210/210 [59:54<00:00, 17.12s/it]
[INFO|trainer.py:3966] 2025-04-16 16:53:08,040 >> Saving model checkpoint to /home/stern/GRPO/offline_rl_v2/output
[INFO|configuration_utils.py:423] 2025-04-16 16:53:08,043 >> Configuration saved in /home/stern/GRPO/offline_rl_v2/output/config.json
[INFO|configuration_utils.py:908] 2025-04-16 16:53:08,044 >> Configuration saved in /home/stern/GRPO/offline_rl_v2/output/generation_config.json
[2025-04-16 16:53:20,865] [INFO] [launch.py:351:main] Process 1558864 exits successfully.
[2025-04-16 16:53:24,869] [INFO] [launch.py:351:main] Process 1558865 exits successfully.
[2025-04-16 16:53:28,874] [INFO] [launch.py:351:main] Process 1558863 exits successfully.
[2025-04-16 16:53:32,878] [INFO] [launch.py:351:main] Process 1558867 exits successfully.
[2025-04-16 16:53:35,882] [INFO] [launch.py:351:main] Process 1558862 exits successfully.
[2025-04-16 16:53:39,886] [INFO] [launch.py:351:main] Process 1558866 exits successfully.
[2025-04-16 16:53:43,890] [INFO] [launch.py:351:main] Process 1558861 exits successfully.
[INFO|modeling_utils.py:3594] 2025-04-16 16:54:36,788 >> The model is bigger than the maximum size per checkpoint (5GB) and is going to be split in 14 checkpoint shards. You can find where each parameters has been saved in the index located at /home/stern/GRPO/offline_rl_v2/output/model.safetensors.index.json.
[INFO|tokenization_utils_base.py:2510] 2025-04-16 16:54:36,791 >> tokenizer config file saved in /home/stern/GRPO/offline_rl_v2/output/tokenizer_config.json
[INFO|tokenization_utils_base.py:2519] 2025-04-16 16:54:36,791 >> Special tokens file saved in /home/stern/GRPO/offline_rl_v2/output/special_tokens_map.json
***** train metrics *****
epoch = 2.968
total_flos = 41570GF
train_loss = 0.0431
train_runtime = 0:59:54.54
train_samples = 2245
train_samples_per_second = 1.874
train_steps_per_second = 0.058
[2025-04-16 16:54:49,957] [INFO] [launch.py:351:main] Process 1558860 exits successfully.
|