Skip to content

Pull requests: AI-Hypercomputer/maxtext

New pull request New
229 Open 4,199 Closed
229 Open 4,199 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

Expose the training mesh on MaxTextTrainingEngine
#4716 opened Aug 3, 2026 by haochengd-google Loading…
4 tasks done
1
Automatically mark all tests as cpu_only unless explicitly tagged with hardware markers
#4715 opened Aug 3, 2026 by SurbhiJainUSC Collaborator • Draft
4 tasks done
1
[MaxText][vLLM] Fix weight converter layout transposes for Qwen3.5 MoE
#4713 opened Aug 3, 2026 by igorts-git Collaborator Loading…
4 tasks done
2
Update example command lines in the module docstring of train_sft.py to remove weight_dtype=bfloat16 argument; this shouldn't be the default weight_dtype for SFT because of potential loss of gradient updates w/master weights in bfloat16.
#4710 opened Aug 3, 2026 by copybara-service Bot Loading…
1
Enable 3D Mrope for Qwen3 in multimodal training
#4709 opened Aug 3, 2026 by lkhphuc Loading…
2 of 4 tasks
1
Fix Qwen3_30b E2E
#4706 opened Aug 3, 2026 by chiajunglien Contributor • Draft
4 tasks
1
Small fix for Qwen3-VL to run vision sft.
#4705 opened Aug 3, 2026 by lkhphuc Loading…
2 of 4 tasks
@hengtaoguo
3
Add Hy3 (Tencent Hunyuan V3) model support
#4704 opened Aug 3, 2026 by weikuo0506 • Draft
7 of 13 tasks
2
Optimize Multi-Token Prediction (MTP) sharding and memory footprint.
#4703 opened Aug 3, 2026 by copybara-service Bot Loading…
@RissyRan
Add myself to more files in codeowners
#4702 opened Aug 3, 2026 by shuningjin Collaborator • Draft
4 tasks done
2
Remove unused checkpoint manager arguments
#4701 opened Aug 3, 2026 by SujeethJinesh Collaborator Loading…
4 tasks done
2
Write post-training checkpoints in MaxText's on-disk layout
#4700 opened Aug 3, 2026 by ecnal-cienet Collaborator • Draft
4 tasks done
2
fix: apply loss cotangent to hidden-state grad in vocab tiling
#4699 opened Aug 2, 2026 by shadowlilac-oss Loading…
3 of 4 tasks
1
Add standalone checkpointer benchmark features and Orbax configuratio…
#4698 opened Jul 31, 2026 by Marlon666 Loading…
1 task
6
Add Pathways health check to run_pathways_tests.yml
#4697 opened Jul 31, 2026 by SurbhiJainUSC Collaborator • Draft
4 tasks done
2
PR permission test, from TPU Raiden team intern, clarify that maxtext…
#4696 opened Jul 31, 2026 by haochengd-google Loading…
1
apple envy moe implementation pull ready
#4695 opened Jul 31, 2026 by entrpn Collaborator Loading…
4 tasks done
5
Update Orbax to 0.12.2 (Pathways Multi-Tiered Checkpointing Support)
#4694 opened Jul 31, 2026 by SujeethJinesh Collaborator • Draft
4 tasks done
2
PR #4640: fix: refine scanned layers unroll logic for Gemma 3
#4693 opened Jul 31, 2026 by copybara-service Bot Loading…
4 tasks done
1
Fix compile_cache_test flakiness caused by the persistent cache compile time threshold
#4692 opened Jul 31, 2026 by bzantium Collaborator Loading…
4 tasks done
2
Validate rampup batch size settings in the Pydantic config
#4691 opened Jul 31, 2026 by Atishyy27 Loading…
1
1
fix(nnx): scope PartitionSpec default strictly to LoRA parameters in sharding analysis
#4689 opened Jul 31, 2026 by copybara-service Bot Loading…
4 tasks done
1
Add Qwen3-VL-2B E2E test scripts
#4688 opened Jul 31, 2026 by chiajunglien Contributor • Draft
4 tasks
1
Add TPU Ulysses context parallelism gemini-review
#4687 opened Jul 31, 2026 by huytransformer Collaborator Loading…
4 tasks
6
Add GPT-OSS-20b E2E test scripts
#4686 opened Jul 31, 2026 by chiajunglien Contributor • Draft
4 tasks done
1
ProTip! Follow long discussions with comments:>50.