-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Pull requests: flashinfer-ai/flashinfer
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix(gdn): compile SM12x CuteDSL kernels as sm_121a on DGX Spark
#3960
opened Jul 14, 2026 by
nv-tusharma
Loading…
3 of 4 tasks
test: enable the unified MoE fuzzer by default (was opt-in)
op: moe
#3958
opened Jul 14, 2026 by
YangXu1990uiuc
Collaborator
Loading…
5 tasks done
feat: expose trtllm-gen block-sparse attention for decode
op: attention
#3955
opened Jul 14, 2026 by
PerkzZheng
Contributor
Loading…
Update Docker CI tags to 20260714-ce40d25
automated
docker
#3954
opened Jul 14, 2026 by
flashinfer-bot
Collaborator
Loading…
fix(gemm): skip low-latency FP8 path for K-major m=32 on SM100
op: gemm
#3953
opened Jul 14, 2026 by
kahyunnam
Member
Loading…
2 tasks
fix(comm): restore MNNVL flags under inference mode
op: comm
run-ci
v0.6.15
#3950
opened Jul 13, 2026 by
galletas1712
Contributor
Loading…
3 tasks done
perf(attention): reuse TRT-LLM-gen counter buffers
op: attention
#3949
opened Jul 13, 2026 by
saltyminty
Collaborator
Loading…
5 tasks done
perf(gemm): update mm_fp4 cute-dsl tactic heuristic and autotune over a top-N ranked list
op: gemm
run-ci
#3948
opened Jul 13, 2026 by
bkryu
Collaborator
Loading…
3 of 5 tasks
perf(jit): drop dead SM12x gencode from SM10x-serving modules
op: gemm
op: moe
#3947
opened Jul 13, 2026 by
yichengj0
Contributor
Loading…
Fix the routing inconsistency for num_groups > 1
op: moe
#3946
opened Jul 13, 2026 by
b8zhong
Contributor
Loading…
Add CuTe DSL HCA backend for DeepSeek V4 sparse MLA
op: attention
#3943
opened Jul 13, 2026 by
myu-guo
Loading…
docs: cover remaining @flashinfer_api symbols missing from rst
#3942
opened Jul 13, 2026 by
kangbintNV
Contributor
Loading…
1 of 3 tasks
docs: add fused_moe LoRA delta APIs to rst
#3941
opened Jul 13, 2026 by
kangbintNV
Contributor
Loading…
BF16 support for the unified MoE fuzzer +
TrtllmBf16Config.prepare_weights
op: moe
#3940
opened Jul 13, 2026 by
feih-nv
Collaborator
Loading…
3 tasks done
Migrate deprecated cutlass.utils spellings to canonical namespaces
op: gemm
op: moe
#3938
opened Jul 13, 2026 by
yiwangchunyu
•
Draft
5 tasks
DeepSeek 4 NVFP4 KV cache model type for sparse_mla on SM_120
op: attention
#3937
opened Jul 12, 2026 by
danielwoz
Loading…
feat(moe): expose pre-reduce buffer from SM90 CUTLASS runner for fused allreduce
op: moe
#3931
opened Jul 12, 2026 by
saidrhs
Loading…
5 tasks done
fix(comm): make find_loaded_library skip look-alike libs like libcudart_stub.so
op: comm
#3930
opened Jul 12, 2026 by
aryanputta
Loading…
Raise at capture time when a default-mode wrapper's run() is captured into a CUDA graph
op: attention
#3928
opened Jul 11, 2026 by
WilsonZheng0327
Loading…
5 tasks done
feat(quantization): support configurable scale-factor layouts for MXFP4
#3927
opened Jul 11, 2026 by
huangzhilin-hzl
Contributor
Loading…
2 of 3 tasks
fix(moe): allocate trtllm-gen expert_weights as bf16 in all launchers
op: moe
run-ci
#3925
opened Jul 11, 2026 by
aleozlx
Collaborator
Loading…
3 of 5 tasks
fix(cute_dsl): migrate off APIs removed in nvidia-cutlass-dsl 4.6 (keeps 4.5.x compat)
op: gemm
#3922
opened Jul 10, 2026 by
YangXu1990uiuc
Collaborator
•
Draft
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.