diff --git a/config.yaml b/config.yaml index a8ea1f5..b6b17de 100644 --- a/config.yaml +++ b/config.yaml @@ -137,7 +137,7 @@ policies: hidden_size: 512 output_size: 512 framestack: 3 - flat_size: 2184 + flat_size: 1928 recurrent: # Assumed to be in the same module as the policy diff --git a/pokemonred_puffer/environment.py b/pokemonred_puffer/environment.py index 4bfade5..fedf7e3 100644 --- a/pokemonred_puffer/environment.py +++ b/pokemonred_puffer/environment.py @@ -516,7 +516,7 @@ def run_action_on_emulator(self, action): self.action_hist[action] += 1 # press button then release after some steps # TODO: Add video saving logic - self.pyboy.send_input(action) + self.pyboy.send_input(VALID_ACTIONS[action]) self.pyboy.send_input(VALID_RELEASE_ACTIONS[action], delay=8) self.pyboy.tick(self.action_freq, render=True) diff --git a/pokemonred_puffer/policies/multi_convolutional.py b/pokemonred_puffer/policies/multi_convolutional.py index 55dd91b..82f588c 100644 --- a/pokemonred_puffer/policies/multi_convolutional.py +++ b/pokemonred_puffer/policies/multi_convolutional.py @@ -27,7 +27,7 @@ def __init__( env, screen_framestack: int = 3, global_map_frame_stack: int = 1, - screen_flat_size: int = 2433, # 14341, + screen_flat_size: int = 1928, # 14341, global_map_flat_size: int = 1600, input_size: int = 512, framestack: int = 1,