Ran Ran

# Description Split Mixtral end-to-end test into 2 scripts, and this PR works with XL ML [PR](https://github.com/GoogleCloudPlatform/ml-auto-solutions/pull/262). # Tests Please describe the tests that you ran on Cloud VM to...

# Description Add 2 more Mixtral tests per [PR](https://github.com/google/maxtext/pull/616)'s request (along with this [PR](https://github.com/GoogleCloudPlatform/ml-auto-solutions/pull/272)): * to generate unscanned ckpt * to run pre-training Note: I was not able to get...

NonMatchingSplitsSizesError on Flax BART with wiki summary dataset

2

### System Info Platform: TPU Python: python3.11 ### Who can help? @sanchit-gandhi ### Information - [X] The official example scripts - [ ] My own modified scripts ### Tasks -...

[DRAFT]: please no review.

# Description - Add matmul implementation to **replace** for loop in MoE models (as a alternative milestone for Megablox due to the blocker summarized [here](https://docs.google.com/document/d/1W5rNemv5D8oqAG0nHIhX8y8mq1VKKGuXDjKV5a6_wX0/edit?tab=t.0#heading=h.xgjl2srtytjt)). Eventually, we want to apply...

# Description We should assert true length of prompt, instead of [padded token length](https://github.com/google/JetStream/blob/eaf0d6ec872944eee981690bb112557a37049ec4/jetstream/engine/token_utils.py#L148-L151) if max_prefill_predict_length less than lengh of prompt. Otherwise, it will provide unexpected output. # Test Run...

[Draft] Add dedicated code owners for codebase

1

# Description Add dedicated code owners for codebase for review. # Tests Expect `This CODEOWNERS file is valid` check. # Checklist Before submitting this PR, please make sure (put X...

# Description Start with a short description of what the PR does and how this is a change from the past. The rest of the description includes relevant details and...