From 2aa6ad6e0b893cdfe40715a855842a2fb041be07 Mon Sep 17 00:00:00 2001 From: Ryan Sullivan Date: Fri, 8 Nov 2024 07:27:24 -0500 Subject: [PATCH] PLR setup but slow --- reinforcement_learning/clean_pufferl.py | 2 +- syllabus_wrapper.py | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/reinforcement_learning/clean_pufferl.py b/reinforcement_learning/clean_pufferl.py index 5dceef8c..cc62d425 100644 --- a/reinforcement_learning/clean_pufferl.py +++ b/reinforcement_learning/clean_pufferl.py @@ -313,7 +313,7 @@ def evaluate(data): next_lstm_state[0][:, env_id], next_lstm_state[1][:, env_id], ) - + print("puffer shape", o.shape) actions, logprob, value, next_lstm_state = data.policy_pool.forwards( o.to(data.device), next_lstm_state ) diff --git a/syllabus_wrapper.py b/syllabus_wrapper.py index 849d7c62..95796fd7 100644 --- a/syllabus_wrapper.py +++ b/syllabus_wrapper.py @@ -105,6 +105,8 @@ def _prepare_state(self, state): new_state.append(np.stack(padded_obs.values())) state = torch.Tensor(np.stack(new_state)).to(self.device) + print("syllabus shape", state.shape) + return state def _set_eval_mode(self):