fix(datasets): use get_from_first_device to safely de-shard tfds payload arrays#366
Open
divyashreepathihalli wants to merge 1 commit intogoogle-deepmind:masterfrom
Open
fix(datasets): use get_from_first_device to safely de-shard tfds payload arrays#366divyashreepathihalli wants to merge 1 commit intogoogle-deepmind:masterfrom
get_from_first_device to safely de-shard tfds payload arrays#366divyashreepathihalli wants to merge 1 commit intogoogle-deepmind:masterfrom
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
This PR fixes a critical crash in acme/datasets/tfds.py introduced during the removal of the deprecated
jax.config.pmap_shmap_mergeflag.Details
Currently, sample_and_postprocess in JaxInMemoryRandomSampleIterator unconditionally calls
x.addressable_shards[0].data.squeeze(0)on all elements of the batched data. This causes anAttributeErrorto be thrown whenever the dataset contains 0-dimensional scalar arrays (which do not possessaddressable_shards) or runs on un-sharded single-device configurations.This PR replaces the hardcoded lambda function with
acme.jax.utils.get_from_first_device(data, as_numpy=False), routing the data through the robust _unreplicate helper added recently. This ensures proper handling ofSingleDeviceSharding, 0-dim scalars, and fully replicated parameters without crashing.Changes:
acme.jax.utilsin tfds.pyjax.tree_util.tree_maplambda call with get_from_first_device to safely strip the pmap replica dimension while preserving non-sharded scalars.