Skip to content

Pull requests: AI-Hypercomputer/maxtext

New pull request New
159 Open 3,657 Closed
159 Open 3,657 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

[WIP-Experimental]Onboard qwen1.5-moe-a2.7b model in MaxText
#4070 opened Jun 4, 2026 by YixuanWang-99 Collaborator Loading…
4 tasks
Implement generic multimodal SFT training pipeline for TFDS.
#4069 opened Jun 4, 2026 by copybara-service Bot Loading…
1
feat(lora): Add Gemma3 weight mappings, vLLM adapter serving, and end…
#4068 opened Jun 4, 2026 by RexBearIU Collaborator Loading…
4 tasks done
1
refactor: streamline LoRA parameter restoration for JAX/NNX models
#4067 opened Jun 4, 2026 by RexBearIU Collaborator Loading…
4 tasks done
1
Fix: Gemma 3 & 4 Base Model and Flax Linen Decoding
#4066 opened Jun 4, 2026 by RexBearIU Collaborator Loading…
4 tasks done
1
Snehalv dsv4 muon
#4065 opened Jun 4, 2026 by snehalv2002 Collaborator • Draft
4 tasks
Migrate distillation MaxTextCheckpointManager to reflect Tunix updated Checkpoint Manager.
#4064 opened Jun 4, 2026 by copybara-service Bot Loading…
1
Add elastic training (Pathways + xpk) guide under docs/
#4063 opened Jun 4, 2026 by inardini Loading…
[WIP] Ragged kernel updates
#4061 opened Jun 4, 2026 by RissyRan Collaborator • Draft
4 tasks
1
Measure routing mismatch for Qwen models
#4060 opened Jun 3, 2026 by xuefgu Collaborator • Draft
4 tasks done
1
Fix cuDNN SDPA autoregressive inference when KV cache batch differs from decode batch
#4058 opened Jun 3, 2026 by sfvaroglu Contributor Loading…
4 tasks done
Remove hardcoded vision hyperparams from Qwen mm preprocessor
#4057 opened Jun 3, 2026 by hengtaoguo Collaborator Loading…
4 tasks done
1
Fix sparse distillation loss and speed up teacher top-k logit saving pull ready
#4056 opened Jun 3, 2026 by ajkv-google Collaborator Loading…
4 tasks done
5
[pallas:sc] Remove use_tc_tiling_on_sc=True, because this is now a default
#4055 opened Jun 3, 2026 by copybara-service Bot Loading…
1
Update vllm/tpu-inference commit and fix vllm installation
#4054 opened Jun 3, 2026 by SurbhiJainUSC Collaborator • Draft
4 tasks done
1
Enable Gemma 4 E2B / E4B inference via vLLM RPA gemini-review
#4053 opened Jun 3, 2026 by gagika Collaborator • Draft
4 tasks done
10
Qwen3 Coder 480B inference sharding changes for MaxText.
#4052 opened Jun 3, 2026 by copybara-service Bot Loading…
1
Support of gdn kernel from tpu-inference gemini-review
#4051 opened Jun 3, 2026 by khatwanimohit Collaborator Loading…
4 tasks done
6
Fix: Refactor LoRA checkpoint restoration and simplify NNX weight extraction
#4050 opened Jun 3, 2026 by RexBearIU Collaborator Loading…
4 tasks done
1
[RL] Honor tokenizer chat templates for base models that lack one
#4049 opened Jun 3, 2026 by dasoto Collaborator Loading…
4 tasks done
2
[WIP-exp1] microsoft/Phi-4-mini-instruct
#4047 opened Jun 3, 2026 by hengtaoguo Collaborator • Draft
4 tasks
1
fix: raise RuntimeError when checkpoint step >= config.steps bug Something isn't working gemini-review pull ready
#4046 opened Jun 2, 2026 by Dr-Left Collaborator Loading…
4 tasks done
@khatwanimohit
7
Add intermediate eval hook: fire evaluate() every eval_interval outer steps
#4044 opened Jun 2, 2026 by py4 Collaborator Loading…
4 tasks done
3
[Qwen3.5] Add moe weight sync script for 35b model gemini-review pull ready
#4041 opened Jun 2, 2026 by Rohan-Bierneni Collaborator Loading…
4 tasks done
@Rohan-Bierneni
6
Support Qwix quantization on NNX
#4040 opened Jun 2, 2026 by hsuan-lun-chiang Collaborator Loading…
4 tasks done
12
ProTip! Type g i on any issue or pull request to go back to the issue listing page.