fix: remove retired jax_enable_async_collective_offload flag from TPU launch script#119
Open
aryanputta wants to merge 1 commit into
Conversation
… launch script The --jax_enable_async_collective_offload flag was retired in recent JAX releases. Accessing it triggers: ERROR: Accessing retired flag 'jax_enable_async_collective_offload' which aborts TPU pod training before the first step. This is the root cause reported in issue young-geng#109. Remove the flag from LIBTPU_INIT_ARGS in examples/pretrain_llama_7b.sh. The async collective offload behaviour it controlled is now managed internally by the XLA/JAX runtime and does not need explicit opt-in. Fixes young-geng#109
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
Running
examples/pretrain_llama_7b.shon TPU pods with recent JAX versions fails immediately with:The flag
--jax_enable_async_collective_offload=trueinLIBTPU_INIT_ARGSwas retired upstream. JAX now manages async collective offload behavior internally and does not accept this flag as an explicit argument.Reported in #109.
Fix
Remove
--jax_enable_async_collective_offload=truefromLIBTPU_INIT_ARGSinexamples/pretrain_llama_7b.sh. All other XLA/TPU flags are retained unchanged.Test
Reproduced the error on a v3 TPU pod with JAX >= 0.4.26. After this change the
ERROR: Accessing retired flagmessage is gone and training proceeds normally.