Skip to content

Pull requests: AI-Hypercomputer/maxtext

New pull request New
275 Open 4,580 Closed
275 Open 4,580 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

Train & Eval on all samples when pdbs < 1
#5158 opened Sep 7, 2026 by muskansh-google Contributor • Draft
4 tasks
3
Fix Multimodal SFT grain pipeline crashes (Tokenizer & IterDataset)
#5156 opened Sep 7, 2026 by chiajunglien Collaborator Loading…
4 tasks done
3
[Stacked PR 4/5] Enable GDN backward pass kernel and hybrid precision in Qwen3 model
#5154 opened Sep 6, 2026 by Rohan-Bierneni Collaborator • 4/5 Loading…
3 tasks done
4
[Stacked PR 3/5] Add decoupled GDN Pallas backward pass kernel and parity tests
#5153 opened Sep 6, 2026 by Rohan-Bierneni Collaborator • 3/5 Loading…
3 tasks done
4
[Stacked PR 2/5] Import local Tokamax GDN forward kernel with custom remat for backward pass support
#5152 opened Sep 6, 2026 by Rohan-Bierneni Collaborator • 2/5 Loading…
3 tasks done
16
[Stacked PR 1/5] Fix upstream pyink formatting and GMM v2 compatibility for Qwen3.5
#5151 opened Sep 6, 2026 by Rohan-Bierneni Collaborator • 1/5 Loading…
3 tasks done
3
Add dense Qwen3.5/3.8-27B checkpoint-conversion support
#5149 opened Sep 4, 2026 by lokic233 Loading…
2
Register LazyTensor with the Orbax v1 leaf-handler registry
#5148 opened Sep 4, 2026 by lokic233 Loading…
1
6
hengtaoguo-exp5
#5146 opened Sep 4, 2026 by hengtaoguo Collaborator • Draft
4 tasks
4
Cl 834
#5145 opened Sep 4, 2026 by muskansh-google Contributor • Draft
7
Trace the engine's eval kernel under the mesh and the axis rules
#5144 opened Sep 4, 2026 by NuojCheng Collaborator Loading…
4 tasks
1
Clip gradients with the norm the engine already computed
#5143 opened Sep 4, 2026 by NuojCheng Collaborator Loading…
4 tasks done
1
Add separate final normalization for Multi-Token Prediction in DeepSeek models.
#5142 opened Sep 4, 2026 by copybara-service Bot Loading…
@CaptainO5
1
Offload MoE collectives and ragged-sort index math to the TPU SparseCore
#5140 opened Sep 4, 2026 by NuojCheng Collaborator • Draft
4 tasks
5
vLLM adapter: index kv_caches through layer_name_to_kvcache_index
#5139 opened Sep 4, 2026 by sierraisland Contributor Loading…
4 tasks done
2
Defer the engine's live-object publish out of the step path
#5138 opened Sep 4, 2026 by NuojCheng Collaborator • Draft
4 tasks
6
Fix AttributeError on missing use_sft in grain data processing
#5137 opened Sep 4, 2026 by bvandermoon Collaborator Loading…
4 tasks done
1
Keep qwix out of tpu-inference's fused MoE kernel and pre-quantize weights
#5136 opened Sep 3, 2026 by sierraisland Contributor Loading…
4 tasks done
4
Internal change
#5135 opened Sep 3, 2026 by copybara-service Bot Loading…
@SurbhiJainUSC
fix(attention test): give the mask tests the axis rules the cp gate reads
#5134 opened Sep 3, 2026 by gulsumgudukbay Collaborator Loading…
4 tasks done
1
fix(mhc test): let the batch scale with the device count
#5133 opened Sep 3, 2026 by gulsumgudukbay Collaborator Loading…
4 tasks done
1
fix(diloco test): set enable_checkpointing explicitly when loading state
#5132 opened Sep 3, 2026 by gulsumgudukbay Collaborator Loading…
4 tasks done
2
fix(nnx): keep only newly created params in _apply_layers_sequentially
#5130 opened Sep 3, 2026 by hodaaaaaaaaaa Loading…
3 of 4 tasks
1
3
feat(kda): integrate KDA attention with tokamax backend and CP support
#5128 opened Sep 3, 2026 by chiaotung97 Collaborator Loading…
4 tasks done
19
Cl 974697834
#5127 opened Sep 3, 2026 by muskansh-google Contributor • Draft
4 tasks
8
ProTip! Updated in the last three days: updated:>2026-09-04.