Skip to content

Pull requests: AI-Hypercomputer/maxtext

New pull request New
237 Open 4,414 Closed
237 Open 4,414 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

Write post-training checkpoints in MaxText's on-disk layout
#4945 opened Aug 19, 2026 by hsuan-lun-chiang Collaborator • Draft
8 tasks done
6
[mHC] Optimize stream mapping with contracted einsums to eliminate 5D HBM roundtrips (+16.9% speedup)
#4944 opened Aug 19, 2026 by copybara-service Bot Loading…
1
[DiLoCo] Add Docs for Streaming DiLoCo
#4942 opened Aug 19, 2026 by Dr-Left Collaborator Loading…
4 tasks done
@Dr-Left
8
fix(e2e): Fix dataset path for multimodal SFT, RL parallelism, and LLaMA to_hf settings
#4941 opened Aug 19, 2026 by RexBearIU Collaborator Loading…
2 tasks done
4
[Gemma4] Fix the bug of use_qk_norm accidently set to False in Gemma 4 configs
#4938 opened Aug 18, 2026 by Dr-Left Collaborator Loading…
4 tasks done
@Dr-Left
8
Fix OOM in to_huggingface conversion and stabilize post-training E2E TPU pipelines
#4937 opened Aug 18, 2026 by SurbhiJainUSC Collaborator • Draft
4 tasks done
3
Add a sharding-aware implementation of muon to Maxtext.
#4936 opened Aug 18, 2026 by copybara-service Bot Loading…
Add support for Tokamax GMM v2 heuristic tiling function.
#4935 opened Aug 18, 2026 by copybara-service Bot Loading…
@CaptainO5
Fix duplicate XLA compilation of jit(train_step)
#4934 opened Aug 18, 2026 by olupton Contributor Loading…
4 tasks done
5
Build the Qwen3-Next shared expert only when shared_experts > 0
#4930 opened Aug 18, 2026 by WandLZhang Loading…
3
Let splash attention exceed 524,288 tokens when packing is off
#4929 opened Aug 18, 2026 by WandLZhang Loading…
3
fix(checkpointing): re-raise fatal checkpointing errors in non-elastic mode
#4927 opened Aug 18, 2026 by RexBearIU Collaborator Loading…
4 tasks done
5
Step0 reforward test scripts
#4925 opened Aug 18, 2026 by entrpn Collaborator Loading…
4 tasks
8
Add HCA Static Compilation for Splash Attention [Deepseek v4]
#4924 opened Aug 17, 2026 by octatrifan Collaborator • Draft
4 tasks done
12
Prepare MaxText release 0.2.4
#4923 opened Aug 17, 2026 by copybara-service Bot Loading…
@SurbhiJainUSC
1
cosmos3_nano model bring up (patch PR#4475)
#4920 opened Aug 17, 2026 by lydhr Collaborator Loading…
4 tasks done
2
Archive older news updates to shorten main list
#4919 opened Aug 17, 2026 by bvandermoon Collaborator Loading…
4 tasks done
1
Add grain_index_storage_option pull ready
#4916 opened Aug 17, 2026 by aireenmei Collaborator Loading…
4 tasks done
6
DSV4: Add indexer_save_selection and indexer_sharded_topk for the CSA indexer
#4915 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
5
DSV4: Add mask hoisting, boolean splash masks, and thresholded indexer membership
#4914 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
MoE: remove the replicated-activation repeat in RoutedMoE.permute
#4913 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
2
DSV4: Add mhc_split_axis_contraction to contract the mHC rate and embed axes separately
#4912 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
3
mesh: add mesh_ring_axis to give a mesh axis physical rings on a torus without wraparound
#4911 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
2
MoE: add moe_late_tp_combine so routed experts take the tensor axis as expert parallelism
#4910 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
2
DSV4: Add shard_indexer_acts to shard the CSA indexer scores across heads
#4909 opened Aug 17, 2026 by systalyze-ai Loading…
4 tasks done
5
ProTip! Follow long discussions with comments:>50.