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

[BugFix] Allow for composite action distributions in PPO/A2C losses #2391

Merged
merged 19 commits into from
Sep 4, 2024
Merged

[BugFix] Allow for composite action distributions in PPO/A2C losses #2391

Changes from 1 commit
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
627b673
account for composite distribution
albertbou92 Aug 10, 2024
9b08555
tests
albertbou92 Aug 11, 2024
a668d0c
fix tests ppo
albertbou92 Aug 12, 2024
e65e1a3
fix tests ppo
albertbou92 Aug 12, 2024
fd60ad7
fix tests ppo
albertbou92 Aug 12, 2024
bc00e2d
fix tests ppo
albertbou92 Aug 12, 2024
e3f61f7
fix tests ppo
albertbou92 Aug 12, 2024
4eb413d
fix tests ppo
albertbou92 Aug 12, 2024
560ddad
fix tests ppo
albertbou92 Aug 12, 2024
2865c92
format
albertbou92 Aug 12, 2024
13a092d
a2c tests
albertbou92 Aug 14, 2024
8bc8e03
a2c tests
albertbou92 Aug 14, 2024
07ec262
a2c tests
albertbou92 Aug 14, 2024
a95eb3d
a2c tests
albertbou92 Aug 14, 2024
8f00828
format
albertbou92 Aug 14, 2024
9ce3c58
docstrings and minor fixes
albertbou92 Aug 14, 2024
ad32b90
docstrings and minor fixes
albertbou92 Aug 14, 2024
37b733f
docstrings fix
albertbou92 Aug 14, 2024
69922fa
added required grad for td
albertbou92 Aug 26, 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
Prev Previous commit
Next Next commit
fix tests ppo
  • Loading branch information
@albertbou92
albertbou92 committed Aug 12, 2024
commit e65e1a3f20b24e38dc7e79de3dce2cad03127ab7
16 changes: 9 additions & 7 deletions test/test_cost.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7750,14 +7750,15 @@ def _create_mock_data_ppo(
},
action_key: {"action1": action} if composite_action_dist else action,
sample_log_prob_key: torch.randn_like(action[..., 1]) / 10,
loc_key: loc,
scale_key: scale,
},
device=device,
)
if composite_action_dist:
td[("params", "action1", loc_key)] = loc
td[("params", "action1", scale_key)] = scale
else:
td[loc_key] = loc
td[scale_key] = scale
return td

def _create_seq_mock_data_ppo(
Expand Down Expand Up @@ -7806,15 +7807,16 @@ def _create_seq_mock_data_ppo(
sample_log_prob_key: (
torch.randn_like(action[..., 1]) / 10
).masked_fill_(~mask, 0.0),
"loc": loc,
"scale": scale,
},
device=device,
names=[None, "time"],
)
if composite_action_dist:
td[("params", "action1", "loc")] = loc
td[("params", "action1", "scale")] = scale
else:
td["loc"] = loc
td["scale"] = scale

return td

Expand Down Expand Up @@ -7882,10 +7884,10 @@ def test_ppo(

loss = loss_fn(td)
if isinstance(loss_fn, KLPENPPOLoss):
if "kl" in loss:
kl = loss.pop("kl")
else:
if composite_action_dist:
kl = loss.pop("kl_approx")
else:
kl = loss.pop("kl")
assert (kl != 0).any()

loss_critic = loss["loss_critic"]
Expand Down