[BugFix] Seed batched native MuJoCo workers distinctly - #4208
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4208
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit 4a30c44 with merge base 573d8dd ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
github-actions
Bot
added
Environments
MujocoEnv(num_envs=N, seed=s, backend="mujoco") built every SerialEnv or ParallelEnv worker from the same kwargs, so all workers shared one RNG stream and drew identical reset noise (and, for subclasses, identical commands). Derive one seed per worker with torchrl._utils.seed_generator, the chain EnvBase.set_seed uses across a batch, and pass it through create_env_kwargs, so each worker gets its own seed while a given seed stays reproducible. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
vmoens
force-pushed
the
mujoco-batched-seed
branch
from
September 3, 2026 10:26
162e29f to
4a30c44
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
MujocoEnv(num_envs=N, seed=s, backend="mujoco")builds one worker per env through_MujocoMetafrom the same kwargs, so everySerialEnv/ParallelEnvworker was seeded withsand replayed the same reset noise (and, forMicroDuckEnv, the same velocity command at every reset). Each worker now receives its own seed throughcreate_env_kwargs, derived from the requested seed withtorchrl._utils.seed_generator(the same chainEnvBase.set_seedwalks across a batch), so the batch stays reproducible for a given seed and a laterset_seedproduces the same assignment.seed=Nonewas unaffected.Found while writing the
MicroDuckTaskdocstring example in #4204: four envs built withseed=0all drew the same command.Types of changes
Checklist
Validation
test_native_batched_workers_draw_distinct_reset_noise(serial and parallel) asserts that two workers built from one seed reset to different observations and that two envs built from the same seed reset identically; it fails onmainand passes with the fix.🤖 Generated with Claude Code