feat: JAX-accelerated normalization-stats pipeline (3.4× faster on ALOHA Pen Uncap; larger gains on bigger datasets) #586
Conversation
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
- Add JAX implementation with order of mangnituted speedup, test on H100/200 GPUs - Two modes: FastRunningStats (default) and JaxRunningStats (exact) - Comprehensive test suite and documentation - Drop-in replacement for original compute_norm_stats.py - Update README with usage instructions Hardware tested: H100/200 GPU
- Add compute_norm_stats_jax.py with GPU acceleration using JAX - Support multi-GPU processing with automatic batch size optimization - Provide FastRunningStats for speed and JaxRunningStats for accuracy - Add comprehensive documentation in docs/jax_acceleration.md - Update README.md with performance information - Remove test file as requested - Fix code quality issues with ruff formatting and linting Performance: 3.4x speedup on ALOHA pen uncap dataset (50k frames) Tested on H100 and current machine with real robotics datasets
marianophielipp
requested review from
Michael-Equi,
jimmyt857,
kvablack and
uzhilinsky
as code owners
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
Adds compute_norm_stats_jax.py, a drop-in, JAX-powered replacement for compute_norm_stats.py.
On the ALOHA Pen Uncap dataset (50k frames) it delivers a 3.4 × speed-up; on larger datasets the gains are even more pronounced, making feasible to train with large number of episodes.