rxnn 0.2.42__py3-none-any.whl → 0.2.43__py3-none-any.whl

rxnn/training/models.py

@@ -168,6 +168,9 @@ class MrlActorModel(nn.Module):
             self.decoder.not_memory_parameters()
         ))
 
+    def embedding_parameters(self) -> Iterator[nn.Parameter]:
+        return self.encoder.model.embedding.parameters()
+
     def unique_parameters(self, with_embedding: bool = True):
         if with_embedding:
             return list(set(

rxnn/training/mrl.py

@@ -225,7 +225,7 @@ class MRLTrainer:
     ) -> tuple[torch.optim.Optimizer, torch.optim.Optimizer]:
         if memory_lr is not None:
             optimizer = torch.optim.AdamW([
-                {'params': self.actor.encoder.embedding.parameters(), 'lr': embedding_lr},
+                {'params': self.actor.embedding_parameters(), 'lr': embedding_lr},
                 {'params': self.actor.not_memory_parameters(), 'lr': lr},
                 {'params': self.actor.memory_parameters(), 'lr': memory_lr},
             ],
@@ -233,7 +233,7 @@ class MRLTrainer:
             )
         else:
             optimizer = torch.optim.AdamW([
-                {'params': self.actor.encoder.embedding.parameters(), 'lr': embedding_lr},
+                {'params': self.actor.embedding_parameters(), 'lr': embedding_lr},
                 {'params': self.actor.unique_parameters(with_embedding=False), 'lr': lr},
             ],
                 weight_decay=weight_decay,
@@ -929,7 +929,7 @@ class MRLTrainer:
 
         if mode == 'update':
             params = [
-                {'params': self.actor.encoder.embedding.parameters(), 'lr': embedding_lr},
+                {'params': self.actor.embedding_parameters(), 'lr': embedding_lr},
                 {'params': self.actor.encoder.not_memory_parameters(), 'lr': model_lr},
                 {'params': self.actor.encoder.memory_parameters(), 'lr': memory_lr},
                 {'params': self.actor.memory_attention_parameters(), 'lr': memory_lr},
@@ -938,7 +938,7 @@ class MRLTrainer:
             ]
         elif mode == 'fetch':
             params = [
-                {'params': self.actor.encoder.embedding.parameters(), 'lr': unfreeze_lr},
+                {'params': self.actor.embedding_parameters(), 'lr': unfreeze_lr},
                 {'params': self.actor.encoder.not_memory_parameters(), 'lr': unfreeze_lr},
                 {'params': self.actor.encoder.memory_parameters(), 'lr': unfreeze_lr},
                 {'params': self.actor.memory_attention_parameters(), 'lr': unfreeze_lr},
@@ -947,7 +947,7 @@ class MRLTrainer:
             ]
         elif mode == 'joint':
             params = [
-                {'params': self.actor.encoder.embedding.parameters(), 'lr': unfreeze_lr},
+                {'params': self.actor.embedding_parameters(), 'lr': unfreeze_lr},
                 {'params': self.actor.encoder.not_memory_parameters(), 'lr': unfreeze_lr},
                 {'params': self.actor.encoder.memory_parameters(), 'lr': memory_lr},
                 {'params': self.actor.memory_attention_parameters(), 'lr': memory_lr},
@@ -956,7 +956,7 @@ class MRLTrainer:
             ]
         else:
             params = [
-                {'params': self.actor.encoder.embedding.parameters(), 'lr': embedding_lr},
+                {'params': self.actor.embedding_parameters(), 'lr': embedding_lr},
                 {'params': self.actor.encoder.not_memory_parameters(), 'lr': model_lr},
                 {'params': self.actor.encoder.memory_parameters(), 'lr': memory_lr},
                 {'params': self.actor.memory_attention_parameters(), 'lr': memory_lr},

{rxnn-0.2.42.dist-info → rxnn-0.2.43.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: rxnn
-Version: 0.2.42
+Version: 0.2.43
 Summary: RxNN: Reactive Neural Networks Platform
 License: Apache-2.0
 Keywords: deep-learning,ai,machine-learning

{rxnn-0.2.42.dist-info → rxnn-0.2.43.dist-info}/RECORD

@@ -16,8 +16,8 @@ rxnn/training/bml.py,sha256=hw6gLpLkGvqLzxIvBg4MvCc5r8cHpEm2RDyh7nH6CtE,16914
 rxnn/training/callbacks.py,sha256=p72lbzFAmFjpcUvyy4aUB3qd53I8C6Sk5w9nQvsKgTk,35852
 rxnn/training/dataset.py,sha256=7hTilFWPpqUEc6zNcMqBPjxFKxCfvTKKF3E8tVlwccQ,51250
 rxnn/training/ddp.py,sha256=VsNBjn3cY-uUj8hbsW7oKvb0_ZKnXnJ2KgObm-Mr9i4,836
-rxnn/training/models.py,sha256=4hDH-R9l1lNvBMW_CGG_QgmCVrkyG7Lyo40PPzvkovQ,8876
-rxnn/training/mrl.py,sha256=xHH-tcmvwmwV5wwiAa3DaXLuF5OipmVDDYxLL5wOYVM,59471
+rxnn/training/models.py,sha256=7IcUsqMdQ3NldXAxHf9mlZw0rmCfSFlRK51nGe4qLAg,8996
+rxnn/training/mrl.py,sha256=4_FYwfI71adrbmDLq9TaBTgpiU8lLAjuR5QYMd57WN4,59423
 rxnn/training/reward.py,sha256=B7nerPk9eNAv2i7umtNF88tVQVwijNNrchIrEITGHKk,11623
 rxnn/training/rl.py,sha256=q4NzIZAmXRHVToT13IHrPTtEikWQUvT0NO0IjApjAO8,6171
 rxnn/training/scheduler.py,sha256=LcjU35mEwz2U5x3U6tLfeeYlBqMxbFSxYzJYuXkWbSY,1408
@@ -33,7 +33,7 @@ rxnn/transformers/moe.py,sha256=j6jEx6Ip0zttlUZKKn82azxo95lkLZs-H2GLSMD88hY,5859
 rxnn/transformers/positional.py,sha256=1PjcJybUzeQlIKJI4tahAGZcYgCRCL0otxs7mpsNuzM,4410
 rxnn/transformers/sampler.py,sha256=t6iiQTdLQ0TakUWnnhKkb5DKF2F_9-thXHBydDF3fxg,17389
 rxnn/utils.py,sha256=ihb6OTyDtPiocB_lOvnq7eOkjjpCkgs8wxvXUBNQ7mM,996
-rxnn-0.2.42.dist-info/LICENSE,sha256=C8coDFIUYuOcke4JLPwTqahQUCyXyGq6WOaigOkx8tY,11275
-rxnn-0.2.42.dist-info/METADATA,sha256=5ZND9je7xzC5qCXQmyFB0XKedtqe5gicSqZnRui1K0Q,25960
-rxnn-0.2.42.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
-rxnn-0.2.42.dist-info/RECORD,,
+rxnn-0.2.43.dist-info/LICENSE,sha256=C8coDFIUYuOcke4JLPwTqahQUCyXyGq6WOaigOkx8tY,11275
+rxnn-0.2.43.dist-info/METADATA,sha256=5k76J0DHE9dpJAteFEPaJO8TYp-D8DrQpD26UeUerJw,25960
+rxnn-0.2.43.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+rxnn-0.2.43.dist-info/RECORD,,