evolutionary-policy-optimization 0.0.46__py3-none-any.whl → 0.0.47__py3-none-any.whl

evolutionary_policy_optimization/epo.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 from functools import partial, wraps
 from pathlib import Path
 from collections import namedtuple
+from random import randrange
 
 import torch
 from torch import nn, cat, stack, is_tensor, tensor
@@ -989,7 +990,8 @@ class EPO(Module):
     @torch.no_grad()
     def forward(
         self,
-        env
+        env,
+        fix_seed_across_latents = True
     ) -> MemoriesAndCumulativeRewards:
 
         self.agent.eval()
@@ -1002,12 +1004,22 @@ class EPO(Module):
 
         for episode_id in tqdm(range(self.episodes_per_latent), desc = 'episode'):
 
+            # maybe fix seed for environment across all latents
+
+            env_reset_kwargs = dict()
+
+            if fix_seed_across_latents:
+                seed = randrange(int(1e6))
+                env_reset_kwargs = dict(seed = seed)
+
+            # for each latent (on a single machine for now)
+
             for latent_id in tqdm(range(self.num_latents), desc = 'latent'):
                 time = 0
 
                 # initial state
 
-                state = env.reset()
+                state = env.reset(**env_reset_kwargs)
 
                 # get latent from pool
 

evolutionary_policy_optimization/mock_env.py

@@ -25,7 +25,8 @@ class Env(Module):
         return self.dummy.device
 
     def reset(
-        self
+        self,
+        seed
     ):
         state = randn(self.state_shape, device = self.device)
         return state

{evolutionary_policy_optimization-0.0.46.dist-info → evolutionary_policy_optimization-0.0.47.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: evolutionary-policy-optimization
-Version: 0.0.46
+Version: 0.0.47
 Summary: EPO - Pytorch
 Project-URL: Homepage, https://pypi.org/project/evolutionary-policy-optimization/
 Project-URL: Repository, https://github.com/lucidrains/evolutionary-policy-optimization

evolutionary_policy_optimization-0.0.47.dist-info/RECORD

@@ -0,0 +1,8 @@
+evolutionary_policy_optimization/__init__.py,sha256=0q0aBuFgWi06MLMD8FiHzBYQ3_W4LYWrwmCtF3u5H2A,201
+evolutionary_policy_optimization/epo.py,sha256=-uRpnD0dKF6h4drVSikm9HnlP2OZ0WYQSWRQcghzd9Y,32242
+evolutionary_policy_optimization/experimental.py,sha256=9FrJGviLESlYysHI3i83efT9g2ZB9ha4u3K9HXN98_w,1100
+evolutionary_policy_optimization/mock_env.py,sha256=202KJ5g57wQvOzhGYzgHfBa7Y2do5uuDvl5kFg5o73g,934
+evolutionary_policy_optimization-0.0.47.dist-info/METADATA,sha256=oSI5NowsOOlQZ5cPmCs-8kYeG6TmzUybpRZt_6-cFWk,6213
+evolutionary_policy_optimization-0.0.47.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+evolutionary_policy_optimization-0.0.47.dist-info/licenses/LICENSE,sha256=1yCiA9b5nhslTavxPjsQAO-wpOnwJR9-l8LTVi7GJuk,1066
+evolutionary_policy_optimization-0.0.47.dist-info/RECORD,,

evolutionary_policy_optimization-0.0.46.dist-info/RECORD

@@ -1,8 +0,0 @@
-evolutionary_policy_optimization/__init__.py,sha256=0q0aBuFgWi06MLMD8FiHzBYQ3_W4LYWrwmCtF3u5H2A,201
-evolutionary_policy_optimization/epo.py,sha256=SAhWgRY8uPQEKFg1_nz1mvh8A6S_sHwnDykhd0F5xEI,31853
-evolutionary_policy_optimization/experimental.py,sha256=9FrJGviLESlYysHI3i83efT9g2ZB9ha4u3K9HXN98_w,1100
-evolutionary_policy_optimization/mock_env.py,sha256=6AIc4mwL_C6JkAxwESJgCLxXHMzCAu2FcffVg3HkSm0,920
-evolutionary_policy_optimization-0.0.46.dist-info/METADATA,sha256=xP2kdKo52-X4Z5XXTPpW0M_NFI0spuigeL7fvqFlsRM,6213
-evolutionary_policy_optimization-0.0.46.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-evolutionary_policy_optimization-0.0.46.dist-info/licenses/LICENSE,sha256=1yCiA9b5nhslTavxPjsQAO-wpOnwJR9-l8LTVi7GJuk,1066
-evolutionary_policy_optimization-0.0.46.dist-info/RECORD,,