Skip to content

Pull requests: NVIDIA/cudnn-frontend

New pull request New
30 Open 290 Closed
30 Open 290 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

Add native SM100 d192/d128 SDPA fprop kernel for DSv3 MLA
#488 opened Aug 5, 2026 by adshen Collaborator Loading…
2 tasks done
8
Add HSTU attention CuTe DSL kernels for Blackwell
#487 opened Aug 5, 2026 by jiayus-nvidia Contributor • Draft
1
1
Add SM120 FROST SDPA backward engine (sdpa_bwd_sm120)
#486 opened Aug 5, 2026 by Adnios Loading…
2 tasks done
17
frost(sdpa): causal right-band widening + per-sequence THD bottom-right diagonal on SM100 cat-enhancements mod-frontend cuDNN frontend APIs, operation graph construction, plans, and user-facing wrappers. orig-nv-eng Reported or requested by NVIDIA engineering.
#485 opened Aug 5, 2026 by vedaanta Collaborator Loading… Frontend 1.27.0
3
frost(sdpa): strict lse/sink/seq-lens execute contract, no torch.empty in execute
#484 opened Aug 5, 2026 by vedaanta Collaborator Loading… Frontend 1.27.0
7
Fix default-stream race: resolve _get_default_stream(None) to torch's current stream cat-enhancements mod-frontend cuDNN frontend APIs, operation graph construction, plans, and user-facing wrappers. orig-nv-eng Reported or requested by NVIDIA engineering.
#483 opened Aug 5, 2026 by vedaanta Collaborator Loading… Frontend 1.27.0
6
Fix absolute paths in installed CMake configs. cat-infra Build, packaging, tooling, dependency, release, or repository maintenance work. mod-frontend cuDNN frontend APIs, operation graph construction, plans, and user-facing wrappers. orig-nv-eng Reported or requested by NVIDIA engineering.
#482 opened Aug 4, 2026 by BillyONeal Loading…
1 of 2 tasks
Frontend 1.27.0
@BillyONeal
10
BSA: add Sage FP8 forward support for Blackwell
#475 opened Aug 4, 2026 by jiayus-nvidia Contributor Loading…
12
https://github.com/HisRicard/cudnn-frontend/pullsTest
#469 opened Aug 3, 2026 by HisRicard • Draft
2 tasks
8
Fix ragged SDPA backward workspace under-allocation for non-token-major stats layouts
#462 opened Jul 31, 2026 by YangXu1990uiuc Collaborator Loading…
5
Support FP32 output and dynamic M in row-scaled FP4 grouped GEMM cat-enhancements mod-cutedsl CuTeDSL kernels, generated kernels, examples, or related integration work. orig-nv-eng Reported or requested by NVIDIA engineering.
#461 opened Jul 31, 2026 by zianglih Contributor Loading…
@zianglih
6
Remove PyTorch from the cutedsl dependency extra
#455 opened Jul 30, 2026 by Anerudhan Collaborator • Draft
1
CSA compressor: review-response fixups for the ratio=128 kernels (follow-up to #427)
#452 opened Jul 30, 2026 by zkyue Contributor Loading…
4
Add MoE + expert-parallel (MoeEp) Python API with MegaMoE CuTe DSL backend
#448 opened Jul 29, 2026 by mhoqueanik Loading…
2 of 6 tasks
14
DSA indexer forward: add a lean SM100 fast path for the head_dim=128 / qhead_per_kv_head=64 regime (up to 1.46x)
#416 opened Jul 21, 2026 by zkyue Contributor Loading…
1
Reject FP8/MXFP8 SDPA forward combinations exposed to a cuDNN 9.24 split-KV bug cat-enhancements mod-frontend cuDNN frontend APIs, operation graph construction, plans, and user-facing wrappers. orig-nv-eng Reported or requested by NVIDIA engineering.
#414 opened Jul 20, 2026 by vedaanta Collaborator Loading… Frontend 1.27.0
3
Add version-adaptive nvvm.atomicrmw wrapper for cutlass-dsl 4.5.x compat
#411 opened Jul 20, 2026 by Anerudhan Collaborator Loading…
1
1
Support zero-element graphs (e.g. SDPA with batch size 0)
#404 opened Jul 17, 2026 by Anerudhan Collaborator • Draft
1
1
Support installing python bindings via cmake --install (fixes #149) cat-ci CI failures, test flakiness, workflow breakage, or automation issues. cat-infra Build, packaging, tooling, dependency, release, or repository maintenance work. mod-infra Infrastructure, CI/CD, build systems, packaging, releases, or repo maintenance. orig-external Reported or requested by an external user, customer, or community contributor.
#403 opened Jul 17, 2026 by Anerudhan Collaborator Loading… Frontend 1.27.0
1
@Anerudhan
4
Add large-tensor convolution fuzzer cat-ci CI failures, test flakiness, workflow breakage, or automation issues. mod-backend cuDNN backend API, graph execution, descriptors, engines, or backend integration. orig-nv-eng Reported or requested by NVIDIA engineering.
#401 opened Jul 16, 2026 by msalasooNV Collaborator Loading… Frontend 1.27.0
@Anerudhan
24
Add FLOOR_MOD (floored modulo) pointwise mode cat-feature Requests for new functionality, APIs, examples, or behavior improvements. mod-backend cuDNN backend API, graph execution, descriptors, engines, or backend integration. orig-nv-eng Reported or requested by NVIDIA engineering.
#393 opened Jul 15, 2026 by vikramchandrashekar • Draft Frontend 1.27.0
@vikramchandrashekar
1
Add MoE + expert-parallel Python API surface, PyTorch reference, and tests
#389 opened Jul 14, 2026 by Anerudhan Collaborator Loading…
8
[DRAFT] Add framework-agnostic operator APIs with JAX support
#363 opened Jul 8, 2026 by mgoldfarb-nvidia • Draft
1
[DRAFT] Jax Support for FE OSS API
#351 opened Jul 6, 2026 by mgoldfarb-nvidia • Draft
13
benchmark: compute causal SDPA FLOP counts without allocating masks
#347 opened Jul 5, 2026 by fallintoplace Contributor Loading…
1
ProTip! Follow long discussions with comments:>50.