Skip to content

Add support for multi-dimensional action spaces to PPO variants #318

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

Merged
araffin merged 11 commits into from
Dec 8, 2025
Merged

Add support for multi-dimensional action spaces to PPO variants #318

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
79997d8
Fix RecurrentPPO and MaskablePPO forward and predict do not reshape a…
immortal-boy Dec 5, 2025
37614b2
Merge branch 'master' into master
araffin Dec 5, 2025
a86bbba
test: add unit test for multi-dimensional action spaces to PPO variants
immortal-boy Dec 6, 2025
b25e58e
reformat tests/test_lstm.py
immortal-boy Dec 6, 2025
701dd5e
make type
immortal-boy Dec 6, 2025
914a104
update changelog
immortal-boy Dec 7, 2025
dee0acd
Merge branch 'master' into master
araffin Dec 8, 2025
a07bb1b
Update test and version
araffin Dec 8, 2025
1372e91
Asserts are not needed anymore
araffin Dec 8, 2025
8bdb74e
Update changelog.rst
araffin Dec 8, 2025
315616c
Fix tests
araffin Dec 8, 2025
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
3 changes: 2 additions & 1 deletion sb3_contrib/common/maskable/policies.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -139,6 +139,7 @@ def forward(
distribution.apply_masking(action_masks)
actions = distribution.get_actions(deterministic=deterministic)
log_prob = distribution.log_prob(actions)
actions = actions.reshape((-1, *self.action_space.shape)) # type: ignore[misc]
return actions, values, log_prob

def extract_features( # type: ignore[override]
Expand Down Expand Up @@ -304,7 +305,7 @@ def predict(
with th.no_grad():
actions = self._predict(obs_tensor, deterministic=deterministic, action_masks=action_masks)
# Convert to numpy
actions = actions.cpu().numpy() # type: ignore[assignment]
actions = actions.cpu().numpy().reshape((-1, *self.action_space.shape)) # type: ignore[assignment]

if isinstance(self.action_space, spaces.Box):
if self.squash_output:
Expand Down
3 changes: 2 additions & 1 deletion sb3_contrib/common/recurrent/policies.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -254,6 +254,7 @@ def forward(
distribution = self._get_action_dist_from_latent(latent_pi)
actions = distribution.get_actions(deterministic=deterministic)
log_prob = distribution.log_prob(actions)
actions = actions.reshape((-1, *self.action_space.shape)) # type: ignore[misc])
return actions, values, log_prob, RNNStates(lstm_states_pi, lstm_states_vf)

def get_distribution(
Expand Down Expand Up @@ -413,7 +414,7 @@ def predict(
states = (states[0].cpu().numpy(), states[1].cpu().numpy())

# Convert to numpy
actions = actions.cpu().numpy()
actions = actions.cpu().numpy().reshape((-1, *self.action_space.shape)) # type: ignore[assignment]

if isinstance(self.action_space, spaces.Box):
if self.squash_output:
Expand Down