Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Modifies loss storage in DDPG, TD3, and Soft Actor Critic #195

Merged
merged 11 commits into from
Aug 4, 2024
Merged

Modifies loss storage in DDPG, TD3, and Soft Actor Critic #195

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
6a9f758
tries to fix issue
prabhatnagarajan Sep 15, 2022
e8be181
using packaging version
prabhatnagarajan Sep 15, 2022
e66d5a0
minor bug fix
prabhatnagarajan Sep 15, 2022
05ed456
updates version to match ours
prabhatnagarajan Sep 15, 2022
776bb8c
merges with upstream/master
prabhatnagarajan Apr 15, 2023
83459b4
Merge remote-tracking branch 'upstream/master'
prabhatnagarajan Dec 28, 2023
b7e6478
detaches losses for records
prabhatnagarajan Apr 12, 2024
b5b35d5
reverts numpy version
prabhatnagarajan Apr 13, 2024
999fe74
Uses tensor.item()
prabhatnagarajan Jul 7, 2024
50bd939
reverts q record computation
prabhatnagarajan Jul 7, 2024
41c0e92
for extends, undoes things
prabhatnagarajan Jul 7, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions pfrl/agents/ddpg.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -168,7 +168,7 @@ def compute_critic_loss(self, batch):
loss = F.mse_loss(target_q, predict_q)

# Update stats
self.critic_loss_record.append(float(loss.detach().cpu().numpy()))
self.critic_loss_record.append(loss.item())

return loss

Expand All @@ -182,7 +182,7 @@ def compute_actor_loss(self, batch):

# Update stats
self.q_record.extend(q.detach().cpu().numpy())
self.actor_loss_record.append(float(loss.detach().cpu().numpy()))
self.actor_loss_record.append(loss.item())

return loss

Expand Down
4 changes: 2 additions & 2 deletions pfrl/agents/soft_actor_critic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -246,8 +246,8 @@ def update_q_func(self, batch):
# Update stats
self.q1_record.extend(predict_q1.detach().cpu().numpy())
self.q2_record.extend(predict_q2.detach().cpu().numpy())
self.q_func1_loss_record.append(float(loss1))
self.q_func2_loss_record.append(float(loss2))
self.q_func1_loss_record.append(loss1.item())
self.q_func2_loss_record.append(loss2.item())

self.q_func1_optimizer.zero_grad()
loss1.backward()
Expand Down
6 changes: 3 additions & 3 deletions pfrl/agents/td3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,8 +213,8 @@ def update_q_func(self, batch):
# Update stats
self.q1_record.extend(predict_q1.detach().cpu().numpy())
self.q2_record.extend(predict_q2.detach().cpu().numpy())
self.q_func1_loss_record.append(float(loss1))
self.q_func2_loss_record.append(float(loss2))
self.q_func1_loss_record.append(loss1.item())
self.q_func2_loss_record.append(loss2.item())

self.q_func1_optimizer.zero_grad()
loss1.backward()
Expand All @@ -241,7 +241,7 @@ def update_policy(self, batch):
# Since we want to maximize Q, loss is negation of Q
loss = -torch.mean(q)

self.policy_loss_record.append(float(loss))
self.policy_loss_record.append(loss.item())
self.policy_optimizer.zero_grad()
loss.backward()
if self.max_grad_norm is not None:
Expand Down