gaspersic/Gymnasium
1
0
Fork 0
mirror of https://github.com/Farama-Foundation/Gymnasium.git synced 2025-08-28 01:07:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

optimize reinforce_invpend_gym_v26.py (#638)

Browse Source
This commit is contained in:
Mohamad Zamini
2023-08-01 09:16:54 -06:00
committed by GitHub
parent f9b9e7fcf4
commit 552bf8c1a7
1 changed files with 14 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
docs/tutorials/training_agents/reinforce_invpend_gym_v26.py
View File

@@ -201,10 +201,14 @@ class REINFORCE:
deltas = torch.tensor(gs) deltas = torch.tensor(gs)
loss = 0 log_probs = torch.stack(self.probs)
# minimize -1 * prob * reward obtained
for log_prob, delta in zip(self.probs, deltas): # Calculate the mean of log probabilities for all actions in the episode
loss += log_prob.mean() * delta * (-1) log_prob_mean = log_probs.mean()
# Update the loss with the mean log probability and deltas
# Now, we compute the correct total loss by taking the sum of the element-wise products.
loss = -torch.sum(log_prob_mean * deltas)
# Update the policy network # Update the policy network
self.optimizer.zero_grad() self.optimizer.zero_grad()
@@ -216,6 +220,7 @@ class REINFORCE:
self.rewards = [] self.rewards = []
# %% # %%
# Now lets train the policy using REINFORCE to master the task of Inverted Pendulum. # Now lets train the policy using REINFORCE to master the task of Inverted Pendulum.
# #