Skip to content

Pull requests: state-spaces/mamba

New pull request New
90 Open 168 Closed
90 Open 168 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

fix: treat non-positive sample temperature as greedy
#1041 opened Sep 4, 2026 by Chessing234 Contributor Loading…
1 task done
fix: honor list/tensor eos_token_id in decode stop
#1040 opened Sep 4, 2026 by Chessing234 Contributor Loading…
1 task done
fix: accept (B, 1, D) in Mamba3.step like Mamba2
#1039 opened Sep 4, 2026 by Chessing234 Contributor Loading…
1 of 2 tasks
[1 of 5] Sequences longer than ~61K tokens silently corrupt Mamba2's chunk state
#1038 opened Sep 3, 2026 by chipsncheese-ie Loading…
Fix SSD minimal initial state shape
#1032 opened Sep 1, 2026 by zjn20030811 Loading…
fix(triton): pass False not None for delta_softplus after manual softplus
#1029 opened Aug 31, 2026 by Chessing234 Contributor Loading…
2 tasks
fix(triton): handle None dt_bias/D in selective_state_update
#1028 opened Aug 31, 2026 by Chessing234 Contributor Loading…
2 tasks
fix(mha): store pre-conv qkv in conv_state during prefill
#1027 opened Aug 31, 2026 by Chessing234 Contributor Loading…
2 tasks
fix: only create CUDA timing events when enable_timing is set
#1026 opened Aug 31, 2026 by Chessing234 Contributor Loading…
2 tasks
fix(mamba3): support GH200 256x256 SISO backward
#1025 opened Aug 29, 2026 by klokedm Loading…
Fix layer_norm backward to reuse the forward dropout mask and scale dx1 by rowscale
#1023 opened Aug 25, 2026 by cyboTiger Loading…
mamba3: access ctx.saved_tensors exactly once in backward
#1021 opened Aug 18, 2026 by jgcb00 Loading…
Add an opt-in Helion kernel backend for the Mamba-2 SSD path
#1012 opened Aug 3, 2026 by shcho1118 Loading…
fix: apply min-p filtering on softmax probabilities
#1010 opened Jul 31, 2026 by Chessing234 Contributor Loading…
2 tasks
fix: MambaLMHeadModel and create_block robustness
#1008 opened Jul 31, 2026 by Chessing234 Contributor Loading…
1 of 2 tasks
test: MambaConfig and varlen helper regression coverage
#1009 opened Jul 31, 2026 by Chessing234 Contributor Loading…
1 task done
fix: harden Hugging Face config/weight loading helpers
#1007 opened Jul 31, 2026 by Chessing234 Contributor Loading…
1 task done
fix: harden setup.py CUDA build guards and wheel fallback messaging
#1004 opened Jul 31, 2026 by Chessing234 Contributor Loading…
1 of 2 tasks
ci: publish wheels for torch 2.11, 2.12.1, and 2.13.0
#1005 opened Jul 31, 2026 by Chessing234 Contributor Loading…
1 of 2 tasks
1
[Triton] Fix triton launch grids under multi gpu scenarios
#1003 opened Jul 30, 2026 by vasqu Contributor Loading…
3
Fix: support TileLang 0.1.9
#1001 opened Jul 29, 2026 by benglewis Loading…
3
Build C++/CUDA extensions with -std=c++20 (torch 2.14 ATen requires C++20)
#1000 opened Jul 24, 2026 by atalman Loading…
3
Fix numerical instability in A_log during training (closes #72)
#996 opened Jul 22, 2026 by KakashiTech Loading…
2
fix: pin apache-tvm-ffi below 0.1.10 for tilelang 0.1.8
#995 opened Jul 18, 2026 by Chessing234 Contributor Loading…
2 tasks
1
fix: force TileLang-contiguous inputs for Mamba3 MIMO
#994 opened Jul 18, 2026 by Chessing234 Contributor Loading…
1 task
1
1
ProTip! Mix and match filters to narrow down what you’re looking for.