rename

dreamer4/dreamer4.py

@@ -1700,20 +1700,20 @@ class DynamicsWorldModel(Module):
 
         return list(set(params) - set(self.video_tokenizer.parameters()))
 
-    def learn_policy_from_generations(
+    def learn_from_experience(
         self,
-        generation: Experience,
+        experience: Experience,
         policy_optim: Optimizer | None = None,
         value_optim: Optimizer | None = None
     ):
-        latents = generation.latents
+        latents = experience.latents
-        actions = generation.actions
+        actions = experience.actions
-        old_log_probs = generation.log_probs
+        old_log_probs = experience.log_probs
-        old_values = generation.values
+        old_values = experience.values
-        rewards = generation.rewards
+        rewards = experience.rewards
 
-        step_size = generation.step_size
+        step_size = experience.step_size
-        agent_index = generation.agent_index
+        agent_index = experience.agent_index
 
         assert all([*map(exists, (old_log_probs, actions, old_values, rewards, step_size))]), 'the generations need to contain the log probs, values, and rewards for policy optimization'
 

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "dreamer4"
-version = "0.0.31"
+version = "0.0.32"
 description = "Dreamer 4"
 authors = [
     { name = "Phil Wang", email = "lucidrains@gmail.com" }

tests/test_dreamer.py

@@ -238,7 +238,7 @@ def test_action_with_world_model():
 
     # take a reinforcement learning step
 
-    actor_loss, critic_loss = dynamics.learn_policy_from_generations(gen)
+    actor_loss, critic_loss = dynamics.learn_from_experience(gen)
 
     actor_loss.backward(retain_graph = True)
     critic_loss.backward()