Skip to content

Pull requests: flashinfer-ai/flashinfer

New pull request New
230 Open 2,085 Closed
230 Open 2,085 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

[feat] add log gate and initial state pool support in blackwell gdn prefill
#3167 opened Apr 24, 2026 by Observer007 Contributor Loading…
4 of 5 tasks
6
Refactor Part 3- Add block-per-token feature in the customized routing method op: moe
#3166 opened Apr 24, 2026 by ChristinaZ Contributor Loading…
5 tasks done
9
fix(gdn): address remaining CodeRabbit feedback from #3001
#3165 opened Apr 24, 2026 by arpera Loading…
5 tasks done
1
fix: remove over-strict K%4 assert in get_shuffle_matrix_sf_a_row_indices run-ci
#3163 opened Apr 24, 2026 by jimmyzho Contributor Loading…
4 tasks
1
3
tests: split test_trtllm_gen_attention.py into prefill / decode / decode-xqa shards op: attention
#3162 opened Apr 23, 2026 by bkryu Collaborator Loading…
5 tasks
@bkryu
8
CICD bug fix: ensure data/ symlinks exist before jit-cache AOT compilation v0.6.10 release blocker label for 0.6.10
#3158 opened Apr 23, 2026 by kahyunnam Member Loading…
5 tasks done
@kahyunnam
11
feat: DiT layer norm fusions for WAN: flashinfer.diffusion_ops model: wan op: misc op: norm run-ci
#3157 opened Apr 23, 2026 by kahyunnam Member Loading…
5 tasks done
@kahyunnam
36
[fix] fix blackwell gdn accuracy issue model: qwen3.5 run-ci
#3156 opened Apr 23, 2026 by Observer007 Contributor Loading…
5 tasks done
17
fix(gdn): use physical SM count for SM100 persistent prefill kernel run-ci
#3155 opened Apr 23, 2026 by arpera Loading…
5 tasks done
11
feat: Add support for LoRa in MOE mxint4 x bf16 trtllm backend op: moe
#3153 opened Apr 23, 2026 by djns99 Contributor • Draft
5 tasks
6
Integrate CUTLASS Small Tile N Blockscaled GEMMs/Grouped GEMMs for SM120 and SM121 op: gemm run-ci
#3152 opened Apr 23, 2026 by depaulmillz Contributor Loading…
5 tasks done
13
fix: pre-convert scale to CUDA tensor in fused_add_rmsnorm_quant bechmark
#3150 opened Apr 22, 2026 by knagaitsev Loading…
5 tasks done
1
5
feat: RMSNorm + RoPE fusion for WAN: flashinfer.diffusion_ops.fused_qk_rmsnorm_rope model: wan op: misc op: norm run-ci
#3148 opened Apr 22, 2026 by kahyunnam Member Loading…
5 tasks done
@kahyunnam
27
Ameyn/wide vec t1
#3147 opened Apr 22, 2026 by ameynaik-hub Contributor • Draft
5 tasks
5
Update CodeOwners
#3146 opened Apr 22, 2026 by dhiraj113 Collaborator Loading…
5 tasks
4
Fix OOB crash in intermediate_states indexing for GDN decode MTP kernel
#3145 opened Apr 22, 2026 by wenscarl Collaborator Loading…
8
improve gdn mtp bf16 state perf for BS<=8 with LDG.128
#3143 opened Apr 22, 2026 by ameynaik-hub Contributor • Draft
5 tasks
8
[codex] add CLI helper for flashinfer-jit-cache wheel installs
#3142 opened Apr 21, 2026 by dierksen Collaborator • Draft
1
3
Add TGV NVFP4 GEMM tactic to mm_fp4 cute-dsl backend (SM100/SM103) op: gemm
#3141 opened Apr 21, 2026 by Sinestro38 Loading…
3 tasks
9
fix(gemm): skip FP4 cuDNN override-shape path on SM120/SM121 (NaN regression from #2910) op: gemm
#3140 opened Apr 21, 2026 by Kh4L Loading…
4
feat(moe-a2a): Update nvlink onesided all-to-all op: comm run-ci
#3139 opened Apr 21, 2026 by trevor-m Contributor Loading…
5 tasks
18
fix(mla): widen page index to int64_t to avoid 32-bit overflow op: attention
#3136 opened Apr 21, 2026 by Tracin Loading…
1 of 5 tasks
11
fix(jit): decouple FLASHINFER_JIT_VERBOSE from debug compilation flags
#3135 opened Apr 21, 2026 by leonardozcm Loading…
4
feat: Add row_starts and dsa_graph_safe to topk op: misc run-ci
#3133 opened Apr 21, 2026 by zianglih Contributor Loading…
5 tasks done
16
feat: Enable FP8 (E4M3/E5M2) in concat_mla_k for optimize long-context prefill performance and refactor type dispatch for BF16/FP16
#3129 opened Apr 21, 2026 by qiching Collaborator Loading…
4 tasks done
8
ProTip! Type g i on any issue or pull request to go back to the issue listing page.