Skip to content

Pull requests: AI-Hypercomputer/maxtext

New pull request New
238 Open 4,489 Closed
238 Open 4,489 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Integrate TE MoEBlock
#5023 opened Aug 26, 2026 by jberchtold-nvidia Contributor • Draft
4 tasks done
3
Dipakg custom qwen3 convergence issue
#5022 opened Aug 26, 2026 by dipakg-lang Collaborator • Draft
4 tasks
1
Fix FP8 custom-gradient state overwrite in NNX training
#5021 opened Aug 26, 2026 by zhenying-liu Contributor Loading…
2
Add dsv3-mlperf-4k custom mesh and rule
#5020 opened Aug 26, 2026 by NuojCheng Collaborator Loading…
4 tasks done
4
feat: Grain FileInstruction pre-extraction & GCS glob optimization
#5019 opened Aug 26, 2026 by abhinavclemson Collaborator Loading…
4 tasks
4
Make three silent failures in the Raiden training path visible pull ready
#5018 opened Aug 26, 2026 by A9isha Collaborator Loading…
4 tasks done
6
Stage all weight sync chunks before listener creation
#5016 opened Aug 26, 2026 by igorts-git Collaborator Loading…
4 tasks done
3
fix(eval mesh): shard attention with the logical axis rules active at eval
#5015 opened Aug 26, 2026 by NuojCheng Collaborator Loading…
4 tasks done
4
Mhc expand
#5014 opened Aug 26, 2026 by muskansh-google Contributor • Draft
4 tasks
6
Fix chat template prompt/completion formatting and SFT masking for Gemma 4 reasoning
#5013 opened Aug 26, 2026 by csgoogle Contributor Loading…
3
Change xpk to cluster toolkit
#5012 opened Aug 26, 2026 by chiajunglien Collaborator • Draft
4 tasks
3
Support on-the-fly FP8 block dequantization in checkpoint utility for Qwen3.5
#5011 opened Aug 26, 2026 by snehalv2002 Collaborator Loading…
4 tasks done
14
Dipakg custom qwen3 0825
#5010 opened Aug 26, 2026 by dipakg-lang Collaborator • Draft
4 tasks
8
perf(qwen3.5): reuse Qwen3-Next's nested block scan
#5009 opened Aug 26, 2026 by NuojCheng Collaborator • Draft
5
Refactor LoRA resharding and modernize Flax NNX variable access
#5008 opened Aug 26, 2026 by RexBearIU Collaborator • Draft
4 tasks done
6
SFT Pipeline for Maxtext-Omni gemini-review
#5005 opened Aug 25, 2026 by subawocit Collaborator Loading…
4 tasks done
17
Upgrade MaxText post-training pinned dependencies to JAX 0.11.1
#5004 opened Aug 25, 2026 by darisoy Collaborator Loading…
7 tasks done
4
Dsv4 mtp support
#5002 opened Aug 25, 2026 by snehalv2002 Collaborator • Draft
4 tasks
7
Fix DeepSeek V4 sliding window attention mask for packed sequences
#5001 opened Aug 25, 2026 by snehalv2002 Collaborator Loading…
4 tasks done
4
Dsv4 param mapping
#5000 opened Aug 25, 2026 by snehalv2002 Collaborator • Draft
4 tasks
8
Onboard explicit sharding to the Gemma models
#4996 opened Aug 25, 2026 by NuojCheng Collaborator Loading…
4 tasks done
12
Test: XLA:TPU host offload fails inside nested lax.scan
#4995 opened Aug 25, 2026 by NuojCheng Collaborator • Draft
4 tasks
3
[Draft] Bharatgen 221
#4994 opened Aug 25, 2026 by muskansh-google Contributor • Draft
4 tasks
5
Mask tiled_lhs on partial-K tile boundaries in _matmul (gmm_v2) so uninitialized/garbage padding rows in LHS don't propagate NaN*0 into the matmul accumulator.
#4993 opened Aug 25, 2026 by copybara-service Bot Loading…
Add explicit sharding support for the Qwen3 model family
#4992 opened Aug 25, 2026 by NuojCheng Collaborator Loading…
4 tasks done
6
ProTip! Type g i on any issue or pull request to go back to the issue listing page.