9/16 19:56发布PyTorch Releasesviable/strict/1789559427: Refine error message for torch.xpu.utilization() on Xe2+ (#197210)PyTorch
9/16 10:32发布PyTorch Releasestrunk/576a6d8db7b7d07d4955b833d7ac56c0500d1e05: [meta] Reject a non-positive pixel_shuffle/unshuffle factor (#195224)PyTorch
9/16 10:32发布PyTorch Releasestrunk/57914c6e94732191855bd7f1da207f16052d5859: Update handler clearance logic (#196102)PyTorch
9/16 19:56发布PyTorch Releasesviable/strict/1789545173: [cuda][nccl] Drop the dead `NCCL_ALLTOALL_SUPPORTED` arm (#196400)PyTorch
9/17 10:58发布PyTorch Releasestrunk/5d1d994be2d895f42bcfda682358b416ceb902c2: [torchfuzz] Stop the matmul family emitting a redundant cast (#197064)PyTorch
9/16 09:53发布PyTorch Releasestrunk/4afcdd7d224e03231a990eb8e62cf23c339d36c3: [Inductor] Preserve fp32 compute promotion in value_expr (#197132)PyTorch
9/2 10:11发布vLLM Releasesv0.29.0rc1: [Bugfix] Handle padded routes in CUTLASS MoE permutations (#54747)vLLM
9/17 16:41发布PyTorch Releasestrunk/71b32515a2bec7d2dc41378630f25646606e8458: Stop double-running inductor tests on ROCm default shards (#197313)PyTorch
9/17 10:58发布PyTorch Releasestrunk/29c701469ca2df4605194613ca70f3bba41e4abe: Enable FA build for SDPA on Windows+CUDA (#186343)PyTorch
9/16 19:56发布PyTorch Releasesviable/strict/1789552248: [MPS] fix rand and uniform_ upper-bound rounding (#197081)PyTorch
9/16 19:56发布PyTorch Releasesviable/strict/1789540965: Update handler clearance logic (#196102)PyTorch
9/16 09:46发布PyTorch Releasestrunk/081f44f105c10ec496471f97634c49227ea9146b: [PyTorch][Inductor] Add GB300 device specifications (#196944)PyTorch
9/17 10:58发布PyTorch Releasestrunk/529cbcaf1575cd70a53fe533176281536bf687fe: [FSDP2] fix crash during unsharding in inference mode (#197327)PyTorch
9/17 10:58发布PyTorch Releasestrunk/5e822a11ee228141bfa86331e9c3a34bdee4b486: [xpu][inductor] Add reduction heuristic for xpu INNER hint (#190986)PyTorch
9/17 10:58发布PyTorch Releasestrunk/20665f569c2701914aa26fc95db15a453c10fa2a: Revert "Add CPython test `test_dataclasses` (#195855)"PyTorch
9/17 10:58发布PyTorch Releasestrunk/4a81d5787ea55c5a1cb73afec4cd30a171f8cb5b: [inductor] Gate remaining CUDA TMA tests on CUDA devices (#197141)PyTorch
9/16 09:02发布PyTorch Releasestrunk/f8d084d5b5ddddc114e3118c077e98d626a304a9: flex_gemm: support block-scaled GEMMs through QUACK (#192839)PyTorch