rxnn 0.2.38__tar.gz → 0.2.39__tar.gz

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.
Files changed (39) hide show
  1. {rxnn-0.2.38 → rxnn-0.2.39}/PKG-INFO +1 -1
  2. {rxnn-0.2.38 → rxnn-0.2.39}/pyproject.toml +1 -1
  3. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/bml.py +0 -8
  4. {rxnn-0.2.38 → rxnn-0.2.39}/LICENSE +0 -0
  5. {rxnn-0.2.38 → rxnn-0.2.39}/README.md +0 -0
  6. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/.DS_Store +0 -0
  7. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/__init__.py +0 -0
  8. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/experimental/__init__.py +0 -0
  9. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/experimental/attention.py +0 -0
  10. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/experimental/models.py +0 -0
  11. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/experimental/moe.py +0 -0
  12. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/memory/__init__.py +0 -0
  13. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/memory/attention.py +0 -0
  14. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/memory/norm.py +0 -0
  15. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/memory/stm.py +0 -0
  16. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/rxt/__init__.py +0 -0
  17. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/rxt/models.py +0 -0
  18. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/__init__.py +0 -0
  19. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/base.py +0 -0
  20. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/callbacks.py +0 -0
  21. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/dataset.py +0 -0
  22. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/ddp.py +0 -0
  23. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/models.py +0 -0
  24. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/mrl.py +0 -0
  25. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/reward.py +0 -0
  26. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/rl.py +0 -0
  27. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/scheduler.py +0 -0
  28. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/tokenizer.py +0 -0
  29. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/training/utils.py +0 -0
  30. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/__init__.py +0 -0
  31. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/attention.py +0 -0
  32. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/ff.py +0 -0
  33. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/layers.py +0 -0
  34. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/mask.py +0 -0
  35. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/models.py +0 -0
  36. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/moe.py +0 -0
  37. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/positional.py +0 -0
  38. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/transformers/sampler.py +0 -0
  39. {rxnn-0.2.38 → rxnn-0.2.39}/src/rxnn/utils.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.3
2
2
  Name: rxnn
3
- Version: 0.2.38
3
+ Version: 0.2.39
4
4
  Summary: RxNN: Reactive Neural Networks Platform
5
5
  License: Apache-2.0
6
6
  Keywords: deep-learning,ai,machine-learning
@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
4
4
 
5
5
  [tool.poetry]
6
6
  name = "rxnn"
7
- version = "0.2.38"
7
+ version = "0.2.39"
8
8
  description = "RxNN: Reactive Neural Networks Platform"
9
9
 
10
10
  license = "Apache-2.0"
@@ -51,10 +51,6 @@ class MLMTrainer(BaseTrainer):
51
51
  model = next(self.model.children()) if isinstance(self.model, DistributedDataParallel) else self.model
52
52
 
53
53
  router_loss = model.encoder.model.moe_router_loss()
54
-
55
- if self.use_ddp:
56
- router_loss = distributed_mean(router_loss)
57
-
58
54
  loss = main_loss + self.moe_aux_loss_scale * router_loss
59
55
 
60
56
  if self.writer is not None:
@@ -156,10 +152,6 @@ class AutoregressiveTrainer(BaseTrainer):
156
152
  model = next(self.model.children()) if isinstance(self.model, DistributedDataParallel) else self.model
157
153
 
158
154
  router_loss = model.model.moe_router_loss()
159
-
160
- if self.use_ddp:
161
- router_loss = distributed_mean(router_loss)
162
-
163
155
  loss = main_loss + self.moe_aux_loss_scale * router_loss
164
156
 
165
157
  if self.writer is not None:
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes
File without changes