project-llm-trainer 0.13.4__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.

This version of project-llm-trainer might be problematic. Click here for more details.

Files changed (32) hide show
  1. llm_trainer/__init__.py +13 -0
  2. llm_trainer/base_trainer.py +707 -0
  3. llm_trainer/checkpoint.py +114 -0
  4. llm_trainer/dataset.py +335 -0
  5. llm_trainer/dpo_trainer.py +311 -0
  6. llm_trainer/ds_checkpoint.py +72 -0
  7. llm_trainer/eval.py +33 -0
  8. llm_trainer/generate_utils.py +463 -0
  9. llm_trainer/grpo_trainer.py +410 -0
  10. llm_trainer/log.py +65 -0
  11. llm_trainer/loss.py +266 -0
  12. llm_trainer/parallel.py +220 -0
  13. llm_trainer/partition_utils.py +219 -0
  14. llm_trainer/ppo_trainer.py +686 -0
  15. llm_trainer/scheduler.py +220 -0
  16. llm_trainer/sft_trainer.py +97 -0
  17. llm_trainer/tokenizer.py +162 -0
  18. llm_trainer/tools.py +116 -0
  19. llm_trainer/train_configs.py +327 -0
  20. llm_trainer/trainer.py +34 -0
  21. llm_trainer/utils.py +630 -0
  22. project_llm_trainer-0.13.4.data/scripts/calc_intermediate_size +15 -0
  23. project_llm_trainer-0.13.4.data/scripts/ddp_train +21 -0
  24. project_llm_trainer-0.13.4.data/scripts/ds_train +17 -0
  25. project_llm_trainer-0.13.4.data/scripts/py_train +12 -0
  26. project_llm_trainer-0.13.4.data/scripts/smart_train +37 -0
  27. project_llm_trainer-0.13.4.data/scripts/vis_log +98 -0
  28. project_llm_trainer-0.13.4.data/scripts/vis_lr +46 -0
  29. project_llm_trainer-0.13.4.dist-info/METADATA +9 -0
  30. project_llm_trainer-0.13.4.dist-info/RECORD +32 -0
  31. project_llm_trainer-0.13.4.dist-info/WHEEL +5 -0
  32. project_llm_trainer-0.13.4.dist-info/top_level.txt +1 -0
llm_trainer/__init__.py ADDED
@@ -0,0 +1,13 @@
1
+ from .trainer import Trainer
2
+ from .sft_trainer import SFTTrainer
3
+ from .dpo_trainer import DPOTrainer
4
+ from .ppo_trainer import PPOTrainer
5
+ from .grpo_trainer import GRPOTrainer
6
+ from .tools import (
7
+ TrainerTools,
8
+ FileDataset,
9
+ estimate_data_size,
10
+ extract_policy_weights_from_ppo,
11
+ extract_value_weights_from_ppo
12
+ )
13
+ from .generate_utils import generate, streaming_generate