Temp testing automation, please ignore#145
Open
RerRayne wants to merge 55 commits intotest_638428288from
Open
Conversation
…tion_matrix_to_predecessor_pointers` methods to probing. PiperOrigin-RevId: 638460408
…s specifications, and add a `track_max_steps` flag to the sampler. This flag will enable or disable length tracking for padding. PiperOrigin-RevId: 638815148
… methods to convert CLRS sample into text. PiperOrigin-RevId: 638980228
…sv with accuracy metric for Gemma 2B, Gemma 2B + RPE and Gemini Flash. PiperOrigin-RevId: 639058968
…ore). PiperOrigin-RevId: 640969794
…y handle Nones. Due to a bug in JAX, JAX previously permitted `jax.tree.map(f, None, x)` where `x` is not `None`, effectively treating `None` as if it were pytree-prefix of any value. But `None` is a pytree container, and it is only a prefix of `None` itself. Fix user code that was relying on this bug. Most commonly, the fix is to write `jax.tree.map(lambda a, b: (None if a is None else f(a, b)), x, y, is_leaf=lambda t: t is None)`. PiperOrigin-RevId: 641687779
Co-authored-by: Olga Kozlova <grenlaykk@gmail.com>
PiperOrigin-RevId: 647327862
… instead of exactly equal sampling. Added a use_hints field so data with and without hints can be identified per sample.
PiperOrigin-RevId: 649473746
Adding huggingface generators for clrs text
PiperOrigin-RevId: 649652992
PiperOrigin-RevId: 649677897
|
Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA). View this failed invocation of the CLA check for more information. For the most up to date status, view the checks section at the bottom of the pull request. |
PiperOrigin-RevId: 650221144
Roll forward PR #104
PiperOrigin-RevId: 651397657
PiperOrigin-RevId: 651482707
PiperOrigin-RevId: 652589410
…kflow PiperOrigin-RevId: 653240239
PiperOrigin-RevId: 653246379
Update pypi-publish.yml
PiperOrigin-RevId: 653667715
…/setup-python@v4`. Also, remove the check for consistency between the package version and release tag, as this is no longer necessary with `actions/checkout@v4`. PiperOrigin-RevId: 653679763
PiperOrigin-RevId: 653690381
This eradicates the protobuf version conflict. PiperOrigin-RevId: 671801866
PiperOrigin-RevId: 677502943
PiperOrigin-RevId: 696161473
PiperOrigin-RevId: 700717268
PiperOrigin-RevId: 707963531
PiperOrigin-RevId: 717526510
…o prevent generating samples that admit multiple valid paths. PiperOrigin-RevId: 717531004
PiperOrigin-RevId: 719711177
The numpy cross product is vectorized, resulting in an extra dimension for the length 1 case. PiperOrigin-RevId: 738834349
PiperOrigin-RevId: 753034846
The `train_lengths` variable was being modified in place for string matching algorithms, affecting subsequent algorithms in the loop. This change introduces `current_algo_train_lengths` to hold algorithm-specific length configurations, preventing unintended side effects. Additionally, unused return values from `make_multi_sampler` are no longer assigned. Was reported in #169 by [@abhitopia ](https://github.com/abhitopia) PiperOrigin-RevId: 819789548
This change prepares for the new `jax.pmap` by implementing the recommended mechanism for accessing the first shard in a sharded array. A common pattern used with `jax.pmap` is to shard an array that is semantically replicated and grabbing the first shard is meant to "unreplicate". However, JAX does not know that a sharded array is actually replicated, so we must now explicitly grab the first shard. The change is under the `jax_pmap_shmap_merge` configuration flag. If `True`, the new `jax.pmap` implementation based on `jax.jit(jax.shard_map)` is used and requires the new explicit shard access. If `False`, the old `jax.pmap` implementation is used and there is a special case in how `x[0]` works. Please see details here: https://docs.jax.dev/en/latest/migrate_pmap.html#int-array-indexing-into-sharded-arrays PiperOrigin-RevId: 850021421
CL/853389459 added `optax.tree.cast_like` which uses `jax.tree.map` internally. This fails when gradients contain None values because JAX no longer considers None as a tree prefix of non-None values. The fix replaces None gradients with actual gradient values before calling optax, then masks the results back to None afterwards. PUBLIC: Fix JAX tree mismatch error in filter_null_grads. PiperOrigin-RevId: 854216449
PiperOrigin-RevId: 854239203
Increase tolerance in `test_hint_loss` from `rtol=1e-4` to `rtol=1e-3` to fix flaky test failures. PiperOrigin-RevId: 864401872
…merge With jax_pmap_shmap_merge=True, the code in nets.py's NetChunked.__call__ was calling addressable_shards on inputs and hints without checking if they are JAX arrays first. When numpy arrays are passed (e.g., during init), this causes AttributeError. This fix adds a robust _get_first helper that: 1. Checks if input is a JAX Array before accessing addressable_shards 2. For numpy arrays, uses simple x[0] indexing 3. For JAX arrays, handles 0-d arrays, SingleDeviceSharding, and properly extracts data from replicated shards PiperOrigin-RevId: 864890502
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
12 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.
No description provided.