[Feature] Isaac Lab 3.x compatibility for wrapper - #4194
Open
theap06 wants to merge 1 commit into
Open
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4194
Note: Links to docs will display an error until the docs builds have been completed.
|
meta-cla
Bot
added
the
CLA Signed
github-actions
Bot
added
BugFix
Environments
theap06
changed the title
theap06
changed the title
theap06
force-pushed
the
isaaclab-v3-compat
branch
from
September 1, 2026 08:59
944fe58 to
a5f1c97
Compare
Collaborator
Author
|
@torchrlbot reviewer @vmoens |
Contributor
Isaac Lab 3.0 mutates the reward buffer in place between steps, so the wrapper now clones the reward alongside terminated and truncated. The tiled-camera path converts pixels through _isaac_data_to_torch, which accepts the v3 warp-backed ProxyArray as well as v2 torch buffers. The per-index reset bridge detects the experimental Warp direct envs (DirectRLEnvWarp), whose _reset_idx takes a boolean mask instead of indices, and converts accordingly. Verified against Isaac Lab 3.0.0-beta2 (kit-less, ovphysx/newton backends, A10): wrap suites pass 8/8 with native_autoreset on Isaac-Cartpole-Direct-v0, Isaac-Velocity-Flat-Anymal-D-v0 and Isaac-Cartpole-Direct-Warp-v0. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hdDvm4HFQe8e7AuG8d72b
theap06
force-pushed
the
isaaclab-v3-compat
branch
from
September 1, 2026 09:07
a5f1c97 to
1198d88
Compare
Collaborator
Author
|
@vmoens will also create an updated isaac lab wrapper for torchRL for their interface. this pr can be merged before the official v3 release as the proxyarray helper works on the v2 tensors already. we also have the mask branch only activates when isaaclab_experimental is installed and the env is a DirectRLEnvWarp. So we are good to merge this beforehand |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Isaac Lab 3.0 (beta2) introduces a factory-based multi-backend architecture. Testing
IsaacLabWrapperagainst it end-to-end (kit-less install,ovphysxbackend, A10) showed the wrapper's step/reset machinery works unmodified in all supported configurations, with two concrete defects fixed here:terminated/truncatedare no longer mutated, but reward now is). The wrapper cloned terminated/truncated but passed reward through as a view of the mutated buffer. It now clones the reward as well.ProxyArrayin the tiled-camera path..data.*buffers are warp-backedProxyArrayobjects in 3.x (with atorchview accessor), sotorch.as_tensoroncamera.data.output[...]no longer yields a usable tensor, and dtype probing returned warp dtypes. A new_isaac_data_to_torchhelper accepts v2torch.Tensorbuffers, v3 proxies, and plain warp arrays (via DLPack).Also adds a docstring note for the 3.x launch API (
launch_simulation/Hydra presets replacingAppLauncher).Results with a basic warp run
'''
env=IsaacLabWrapper(env=<OrderEnforcing<CartpoleWarpEnv>>, batch_size=torch.Size([8]))
batch_size: torch.Size([8]) device: cuda:0
rollout shape: torch.Size([8, 50])
reward sum: 94.55167388916016
done fraction: 0.03999999910593033
episode_reward: tensor([ 2.7660, 3.7561, 15.9770, 1.4488, 4.8248, 11.1499, 3.3809, 0.6895])
'''