evolutionary-policy-optimization 0.0.40__tar.gz → 0.0.41__tar.gz

{evolutionary_policy_optimization-0.0.40 → evolutionary_policy_optimization-0.0.41}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: evolutionary-policy-optimization
-Version: 0.0.40
+Version: 0.0.41
 Summary: EPO - Pytorch
 Project-URL: Homepage, https://pypi.org/project/evolutionary-policy-optimization/
 Project-URL: Repository, https://github.com/lucidrains/evolutionary-policy-optimization
@@ -162,4 +162,25 @@ agent.load('./agent.pt')
 }
 ```
 
+```bibtex
+@inproceedings{Khadka2018EvolutionGuidedPG,
+    title   = {Evolution-Guided Policy Gradient in Reinforcement Learning},
+    author  = {Shauharda Khadka and Kagan Tumer},
+    booktitle = {Neural Information Processing Systems},
+    year    = {2018},
+    url     = {https://api.semanticscholar.org/CorpusID:53096951}
+}
+```
+
+```bibtex
+@article{Fortunato2017NoisyNF,
+    title   = {Noisy Networks for Exploration},
+    author  = {Meire Fortunato and Mohammad Gheshlaghi Azar and Bilal Piot and Jacob Menick and Ian Osband and Alex Graves and Vlad Mnih and R{\'e}mi Munos and Demis Hassabis and Olivier Pietquin and Charles Blundell and Shane Legg},
+    journal = {ArXiv},
+    year    = {2017},
+    volume  = {abs/1706.10295},
+    url     = {https://api.semanticscholar.org/CorpusID:5176587}
+}
+```
+
 *Evolution is cleverer than you are.* - Leslie Orgel

{evolutionary_policy_optimization-0.0.40 → evolutionary_policy_optimization-0.0.41}/README.md

@@ -110,4 +110,25 @@ agent.load('./agent.pt')
 }
 ```
 
+```bibtex
+@inproceedings{Khadka2018EvolutionGuidedPG,
+    title   = {Evolution-Guided Policy Gradient in Reinforcement Learning},
+    author  = {Shauharda Khadka and Kagan Tumer},
+    booktitle = {Neural Information Processing Systems},
+    year    = {2018},
+    url     = {https://api.semanticscholar.org/CorpusID:53096951}
+}
+```
+
+```bibtex
+@article{Fortunato2017NoisyNF,
+    title   = {Noisy Networks for Exploration},
+    author  = {Meire Fortunato and Mohammad Gheshlaghi Azar and Bilal Piot and Jacob Menick and Ian Osband and Alex Graves and Vlad Mnih and R{\'e}mi Munos and Demis Hassabis and Olivier Pietquin and Charles Blundell and Shane Legg},
+    journal = {ArXiv},
+    year    = {2017},
+    volume  = {abs/1706.10295},
+    url     = {https://api.semanticscholar.org/CorpusID:5176587}
+}
+```
+
 *Evolution is cleverer than you are.* - Leslie Orgel

{evolutionary_policy_optimization-0.0.40 → evolutionary_policy_optimization-0.0.41}/evolutionary_policy_optimization/epo.py

@@ -798,6 +798,10 @@ class Agent(Module):
 
                 latents = self.latent_gene_pool(latent_id = latent_gene_ids)
 
+                orig_latents = latents
+                latents = latents.detach()
+                latents.requires_grad_()
+
                 # learn actor
 
                 logits = self.actor(states, latents)
@@ -822,6 +826,14 @@ class Agent(Module):
                 self.critic_optim.step()
                 self.critic_optim.zero_grad()
 
+                # maybe update latents, if not frozen
+
+                if not self.latent_gene_pool.frozen_latents:
+                    orig_latents.backward(latents.grad)
+
+                    self.latent_optim.step()
+                    self.latent_optim.zero_grad()
+
         # apply evolution
 
         self.latent_gene_pool.genetic_algorithm_step(fitness_scores)
@@ -951,7 +963,7 @@ class EPO(Module):
 
                 done = tensor(False)
 
-                while time < self.max_episode_length:
+                while time < self.max_episode_length and not done:
 
                     # sample action
 

{evolutionary_policy_optimization-0.0.40 → evolutionary_policy_optimization-0.0.41}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "evolutionary-policy-optimization"
-version = "0.0.40"
+version = "0.0.41"
 description = "EPO - Pytorch"
 authors = [
     { name = "Phil Wang", email = "lucidrains@gmail.com" }