log prob calculated after data collected

seungeunrho · Nov 20, 2019 · efd145e · efd145e
1 parent 36d0f5b
commit efd145e
Showing 1 changed file with 3 additions and 3 deletions.
diff --git a/REINFORCE.py b/REINFORCE.py
@@ -28,9  28,9 @@ def put_data(self, item):
 
     def train_net(self):
         R = 0
-        for r, log_prob in self.data[::-1]:
         for r, prob in self.data[::-1]:
             R = r   gamma * R
-            loss = -log_prob * R
             loss = -torch.log(prob) * R
             self.optimizer.zero_grad()
             loss.backward()
             self.optimizer.step()
@@ -49,7  49,7 @@ def main():
             m = Categorical(prob)
             a = m.sample()
             s_prime, r, done, info = env.step(a.item())
-            pi.put_data((r,torch.log(prob[a])))
             pi.put_data((r,prob[a]))
 
             s = s_prime
             score  = r