Latest releases

GitHub releases from top ML projects, refreshed on a schedule.
Project Release Summary
October 24, 2026 02:22 UTC
hip: fix bf16 WMMA operand vector type for gfx11/gfx12 ( #30885 ) The __builtin_amdgcn_wmma_f32_16x16x16_bf16_w32 (and _gfx12) builtins take their A/B operands as vectors of short…
October 24, 2026 01:32 UTC
Fixes #198486 Supersedes #198489 . That PR was closed while the issue awaited maintainer direction. A PT2 collaborator has since invited this PR in #198486 . This resubmission con…
October 24, 2026 01:30 UTC
Test windows AMI - don't merge
October 24, 2026 01:24 UTC
Mark the format stable; fields are unchanged from v0.2. Test Plan: python test/torchci/test_testing.py -k TestReportJsonl -k TestReportEnablement -k TestReportCrashes -k TestRepor…
October 24, 2026 01:23 UTC
Summary We're done testing in the fortesting database, so start writing the owners to the real database. Pull Request resolved: #201911 Approved by: https://github.com/georgehong …
October 24, 2026 01:21 UTC
Bumps setuptools from 78.1.1 to 84.1.0. Release notes Changelog Commits updated-dependencies: dependency-name: setuptools dependency-version: 84.1.0 dependency-type: direct:produc…
October 24, 2026 01:04 UTC
[ghstack-poisoned]
October 24, 2026 01:04 UTC
[ghstack-poisoned]
October 24, 2026 00:35 UTC
[ghstack-poisoned]
October 24, 2026 00:35 UTC
[ghstack-poisoned]
October 24, 2026 00:35 UTC
update vision commit hash
October 24, 2026 00:33 UTC
update vllm commit hash
October 24, 2026 00:33 UTC
update torchtitan commit hash
October 24, 2026 00:33 UTC
update torchtitan commit hash
October 24, 2026 00:13 UTC
[AI Codemod][PerfAICT] Skip TLS lookup in c10::memoryProfilingEnabled…
October 24, 2026 00:13 UTC
map_block() reserves a virtual address (VA) range with cuMemAddressReserve. No code frees this range. ~AllocationRef unmaps and releases the memory, but it does not call cuMemAddr…
October 24, 2026 00:13 UTC
map_block() reserves a virtual address (VA) range with cuMemAddressReserve. No code frees this range. ~AllocationRef unmaps and releases the memory, but it does not call cuMemAddr…
October 24, 2026 00:05 UTC
ggm-cuda: Enforce DP4A MMQ tile height multiple of warp size ( #30837 ) Added static assertion to enforce DP4A MMQ tile height requirement. Website: https://llama.app Attestations…
October 24, 2026 00:04 UTC
[ghstack-poisoned]
October 24, 2026 00:04 UTC
[ghstack-poisoned]