project-llm-trainer 0.8.3__py3-none-any.whl → 0.9.0__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.

@@ -1,5 +1,4 @@
1
1
  from typing import Union, Optional, List
2
- from contextlib import nullcontext
3
2
  import torch
4
3
  from llm_model import VlmModel, KVCache
5
4
  from .tools import TrainerTools
@@ -123,7 +123,7 @@ class OptimConfig:
123
123
  class LossConfig:
124
124
  critical_tokens: Optional[List[int]] = None
125
125
  critical_alpha: float = 1.0
126
- aux_loss_coef: Optional[float] = 1.0
126
+ aux_loss_coef: Optional[float] = 0.001
127
127
 
128
128
 
129
129
  @dataclass(kw_only=True)
@@ -35,5 +35,3 @@ if __name__ == '__main__':
35
35
 
36
36
  plt.plot(steps, losses)
37
37
  plt.show()
38
-
39
-
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: project_llm_trainer
3
- Version: 0.8.3
3
+ Version: 0.9.0
4
4
  Summary: LLM and VLM trainer
5
5
  Author: qibin
6
6
  Author-email: qibin0506@gmail.com
@@ -4,7 +4,7 @@ llm_trainer/dataset.py,sha256=UL3fGeM4XSlyNQRZH-139u3LujqAQx3YyaxNRewk6LE,8935
4
4
  llm_trainer/dpo_trainer.py,sha256=Qi7WKhFO4fdnj9W8BNIF_so6-F8g_YKUoPU9sNjWK_M,13320
5
5
  llm_trainer/ds_checkpoint.py,sha256=X2IWgpgi0yOtogph7n6DEwvK_0Ceb7juu1WMutv3HSk,2270
6
6
  llm_trainer/eval.py,sha256=ZyUfSo2Q8P-lrCdPEnGkoo5pGubd0AabREK5eMISRII,1109
7
- llm_trainer/generate_utils.py,sha256=8K3YFbp7IF_lCkmkzjHhqTW26EBFb2AilQmarVcfMvs,15001
7
+ llm_trainer/generate_utils.py,sha256=9-F29aWKDnAnnDCU_D8uKUkVcM2r6xktwyqV4nlggTg,14966
8
8
  llm_trainer/grpo_trainer.py,sha256=3CcV-cuyV4ZUTymN9vz3au4uf3gZdyo8SGgSj2NEofs,16774
9
9
  llm_trainer/log.py,sha256=XwychwKF6gvFPhthCIZCAEUZ0G3DY3fiQrOHqPWsxz0,463
10
10
  llm_trainer/loss.py,sha256=RhTxftLMj1Tqc5pkUvJiZumfbMEPWL8GBGxdTfQggmk,6744
@@ -17,17 +17,17 @@ llm_trainer/scheduler.py,sha256=LAI_0VxClsIQkix0bRoduRD4vPfVuIZDhZgTAT_KK8k,4901
17
17
  llm_trainer/sft_trainer.py,sha256=rSOGZx53jMgOuJdztfxQASYJ62uD0dVaih4IAnSwGBc,1787
18
18
  llm_trainer/tokenizer.py,sha256=0-xQCMz1xiPTDAZiYsVsiECSoZ_1eIvW9XsZOoFfakQ,7250
19
19
  llm_trainer/tools.py,sha256=5op5qrjjkK-Lr9oes5VxIVnOVYOYGoAdlIJq9mPUf64,2637
20
- llm_trainer/train_configs.py,sha256=pPZkbliRdTnWSv3TUuTM23x9RDdMhGSPrxbNAyzDklY,7636
20
+ llm_trainer/train_configs.py,sha256=u3b3RhZn7S6vm2DKNF5sNRHzp73-eYnActGSlzcIbzs,7638
21
21
  llm_trainer/trainer.py,sha256=sz350VaLR7odGm0vYGTGvFWcr5bC7tmwMx7PUY6LA2o,28980
22
22
  llm_trainer/utils.py,sha256=xC5plG-8-_Al5yIF5xIU5lroOcBBk98TEhtUJrazZPE,12305
23
- project_llm_trainer-0.8.3.data/scripts/calc_intermediate_size,sha256=AggpgNHokJiJMbEtVdOnolqr_4bH3i1UYuZNEAzC2Gc,460
24
- project_llm_trainer-0.8.3.data/scripts/ddp_train,sha256=eZSud6KYQAoKLsYB5QB-FI2zq5AZm6Apq1azKdupV3o,477
25
- project_llm_trainer-0.8.3.data/scripts/ds_train,sha256=41q4rOxwbvZDUY0FDdAIpG13PEaUWBpthhvFvww8uOc,388
26
- project_llm_trainer-0.8.3.data/scripts/plot_loss,sha256=MzFcdJESlVr1srj4Td6-AxPGUKkfB_QEcJwm0Bd-5fU,910
27
- project_llm_trainer-0.8.3.data/scripts/plot_lr,sha256=w_7XR_x3KYYyboeOVAeu_I4fveLFI-C0wBmRrNlmWUI,894
28
- project_llm_trainer-0.8.3.data/scripts/py_train,sha256=tOp9TquORQeU8XN5H7OVIk5O0Ypwi34p_GENxTwgwdk,265
29
- project_llm_trainer-0.8.3.data/scripts/smart_train,sha256=N8dp2n7k6bghGczedBVwOdtf1O66oM_cNPh9QmZt0bM,914
30
- project_llm_trainer-0.8.3.dist-info/METADATA,sha256=WxVWmIDlcVdlNbOIglZo7asnIQDmfvobKEFiloTvjrs,195
31
- project_llm_trainer-0.8.3.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
32
- project_llm_trainer-0.8.3.dist-info/top_level.txt,sha256=LtRFg28i0QIG7iBCD2t095oSco99LCtkijibS9cMGik,12
33
- project_llm_trainer-0.8.3.dist-info/RECORD,,
23
+ project_llm_trainer-0.9.0.data/scripts/calc_intermediate_size,sha256=AggpgNHokJiJMbEtVdOnolqr_4bH3i1UYuZNEAzC2Gc,460
24
+ project_llm_trainer-0.9.0.data/scripts/ddp_train,sha256=eZSud6KYQAoKLsYB5QB-FI2zq5AZm6Apq1azKdupV3o,477
25
+ project_llm_trainer-0.9.0.data/scripts/ds_train,sha256=41q4rOxwbvZDUY0FDdAIpG13PEaUWBpthhvFvww8uOc,388
26
+ project_llm_trainer-0.9.0.data/scripts/plot_loss,sha256=O9ooioAJ-79-X06LosgqF8XOqQe-beRxYm3LsLunmoU,908
27
+ project_llm_trainer-0.9.0.data/scripts/plot_lr,sha256=w_7XR_x3KYYyboeOVAeu_I4fveLFI-C0wBmRrNlmWUI,894
28
+ project_llm_trainer-0.9.0.data/scripts/py_train,sha256=tOp9TquORQeU8XN5H7OVIk5O0Ypwi34p_GENxTwgwdk,265
29
+ project_llm_trainer-0.9.0.data/scripts/smart_train,sha256=N8dp2n7k6bghGczedBVwOdtf1O66oM_cNPh9QmZt0bM,914
30
+ project_llm_trainer-0.9.0.dist-info/METADATA,sha256=PSaXkuGCDIwRi8gPHD3LEYXmDhmMj9rC_8xlZP5ftxw,195
31
+ project_llm_trainer-0.9.0.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
32
+ project_llm_trainer-0.9.0.dist-info/top_level.txt,sha256=LtRFg28i0QIG7iBCD2t095oSco99LCtkijibS9cMGik,12
33
+ project_llm_trainer-0.9.0.dist-info/RECORD,,