homa 0.3.13__py3-none-any.whl → 0.3.15__py3-none-any.whl

homa/rl/sac/SoftActor.py

@@ -29,6 +29,7 @@ class SoftActor:
         )
 
     def train(self, states: torch.Tensor, critic_network: torch.nn.Module):
+        self.network.train()
         self.optimizer.zero_grad()
         loss = self.loss(states=states, critic_network=critic_network)
         loss.backward()
@@ -64,6 +65,6 @@ class SoftActor:
         action = torch.tanh(pre_tanh)
 
         probabilities = distribution.log_prob(pre_tanh).sum(dim=1, keepdim=True)
-        correction = torch.log(1 - action.pow(2) + 1e-6).sum(dim=1, keepdim=True)
+        probabilities -= torch.log(1 - action.pow(2) + 1e-6).sum(dim=1, keepdim=True)
 
-        return action, probabilities - correction
+        return action, probabilities

homa/rl/sac/SoftCritic.py

@@ -48,6 +48,7 @@ class SoftCritic:
         next_states: torch.Tensor,
         actor: SoftActor,
     ):
+        self.network.train()
         self.optimizer.zero_grad()
         loss = self.loss(
             states=states,
@@ -94,5 +95,5 @@ class SoftCritic:
         return rewards + self.gamma * termination_mask * entropy_q
 
     def update(self):
-        soft_update(network=self.network.alpha, target=self.target.alpha)
-        soft_update(network=self.network.beta, target=self.target.beta)
+        soft_update(network=self.network.alpha, target=self.target.alpha, tau=self.tau)
+        soft_update(network=self.network.beta, target=self.target.beta, tau=self.tau)

homa/rl/utils.py

@@ -2,6 +2,6 @@ import torch
 
 
 @torch.no_grad()
-def soft_update(self, network: torch.nn.Module, target: torch.nn.Module):
+def soft_update(network: torch.nn.Module, target: torch.nn.Module, tau: float):
     for s, t in zip(network.parameters(), target.parameters()):
-        t.data.copy_(self.tau * s.data + (1 - self.tau) * t.data)
+        t.data.copy_(tau * s.data + (1 - tau) * t.data)

{homa-0.3.13.dist-info → homa-0.3.15.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: homa
-Version: 0.3.13
+Version: 0.3.15
 Summary: A curated list of machine learning and deep learning helpers.
 Author-email: Taha Shieenavaz <tahashieenavaz@gmail.com>
 Requires-Python: >=3.7

{homa-0.3.13.dist-info → homa-0.3.15.dist-info}/RECORD

@@ -101,7 +101,7 @@ homa/rl/DRQN.py,sha256=zooojji9aeeubOP7cRPSHg31u2Assxk-qjXyGUWIO3A,49
 homa/rl/DiversityIsAllYouNeed.py,sha256=8yKzlVdLisForGyXqxaXUAWG_dozq7dNY8MBasCvniE,3322
 homa/rl/SoftActorCritic.py,sha256=N8EsiYbsLH-dpT2EmqdYFG9KvHNfO3JX8SG2LPTy94s,1962
 homa/rl/__init__.py,sha256=EaNDkIzLH1Oy0Wc0aAyyVs4HVMcZS1tdHDh631LKSXs,146
-homa/rl/utils.py,sha256=ySNGwWFBgN3Phg7Xn99CIJVHRWw3AmU9KcY-GP4ZQBc,236
+homa/rl/utils.py,sha256=IqbN5aDLwovocpPbxgywuetjz7GQwh9aJ4WFIOtLP3g,232
 homa/rl/buffers/Buffer.py,sha256=YCESh9tFxgWOLzGQj_IA0zLJoZWDmz6gCNu1iYsGp1s,388
 homa/rl/buffers/DiversityIsAllYouNeedBuffer.py,sha256=Nwcqs3Q10x6OKZ-zWug4IcBc6RR1TwEIybuFQOtmftA,1612
 homa/rl/buffers/ImageBuffer.py,sha256=HSmMt82hmkL3ooBYo7c6YUtTsMz9TAA8CvPh3y8z3yg,65
@@ -118,8 +118,8 @@ homa/rl/diayn/modules/ContinuousActorModule.py,sha256=yeC117I5gkXZSidQhjwakjiY7G
 homa/rl/diayn/modules/CriticModule.py,sha256=OUenwCG0dG4PnK7Iq-jy7oCTv_Cn9s7bXRpro6Pvb40,956
 homa/rl/diayn/modules/DiscriminatorModule.py,sha256=D58dKBv4f6gtrpqMKLK8XAZpiMqKfS4sG6s3QcF8iGE,891
 homa/rl/diayn/modules/__init__.py,sha256=1Pgjr4FT5WG-AMh26NPEfbf5pK6I02B1x8HYsgyUCJ4,149
-homa/rl/sac/SoftActor.py,sha256=CxR58IFrZ6xlmBj_gq_abZfgdzlVD71c6wA6wQiVL2c,2142
-homa/rl/sac/SoftCritic.py,sha256=EOX1vpH7YVwDuy-RdFgEpIKGioE7si0awoAYrHMTv4g,2923
+homa/rl/sac/SoftActor.py,sha256=NSTqnv_BZzTqfgEEOIEtOgYV2_VycicIF0alD1O5Nk8,2162
+homa/rl/sac/SoftCritic.py,sha256=rOgPR8zRUtjEwF9W4q5nZQaGXFmf_9tmXqaRWzUkAm8,2980
 homa/rl/sac/__init__.py,sha256=8EIkOcVvxN94gGzcZoX2XTnvTsHqW6yBaZ2RdFwIveM,68
 homa/rl/sac/modules/DualSoftCriticModule.py,sha256=Ax28i7U-KnP4QJig-AeeCfpPYNvTT3DfvRMJI-f-TGY,749
 homa/rl/sac/modules/SoftActorModule.py,sha256=LQ4z7s8mE3wwb1JgxPs0QvnriZULK3_ULdhkt60Ffpw,1152
@@ -143,8 +143,8 @@ homa/vision/concerns/__init__.py,sha256=mrw1YvN-GpQPvMwDF00KxnFkksPKo23RWM4KRioU
 homa/vision/modules/ResnetModule.py,sha256=eFudBnILD6OmgQtcW_CQQ8aZ62NEa4HyZ15-lobTtt0,712
 homa/vision/modules/SwinModule.py,sha256=3ZtUcfyJt0NMGmIlGpN35MIJG9QsgcLdFniZH7NxZQo,1227
 homa/vision/modules/__init__.py,sha256=zVMYB9IAO_xZylC1-N3p8ymHgEkAE2sBbuVz8K5Y1kk,74
-homa-0.3.13.dist-info/METADATA,sha256=1oENQJVWWN-hpkF440JOK0IiOI-i-Vbx9zcl7y5_pcs,1760
-homa-0.3.13.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-homa-0.3.13.dist-info/entry_points.txt,sha256=tJZzjs-f2QvFe3ES8Qta8IE5sAbeE8-cyZ_UtbgqG4s,51
-homa-0.3.13.dist-info/top_level.txt,sha256=tmOfy2tuaAwc3W5-i6j61_vYJsXgR4ivBWkhJ3ZtJDc,5
-homa-0.3.13.dist-info/RECORD,,
+homa-0.3.15.dist-info/METADATA,sha256=jo-jsI9A6KvK95nGmxrh_IExb1WeNp_3DWTZgMDSxcI,1760
+homa-0.3.15.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+homa-0.3.15.dist-info/entry_points.txt,sha256=tJZzjs-f2QvFe3ES8Qta8IE5sAbeE8-cyZ_UtbgqG4s,51
+homa-0.3.15.dist-info/top_level.txt,sha256=tmOfy2tuaAwc3W5-i6j61_vYJsXgR4ivBWkhJ3ZtJDc,5
+homa-0.3.15.dist-info/RECORD,,