Commits

Commit history

1. Commits on May 12, 2025

   • Revert "[ROCm] Maxpool forward NHWC Perf Improvement targeting Resnet scenarios (#151727)"

     Show description for daca611
     pytorchmergebot

     committed
     daca611View commit details

     Copy full SHA for daca611

     View code at this pointBrowse repository at this point

   • Revert "Forward fix #151727 (#153306)"

     Show description for 8511d21
     pytorchmergebot

     committed
     8511d21View commit details

     Copy full SHA for 8511d21

     View code at this pointBrowse repository at this point
2. Commits on May 11, 2025

   • Forward fix #151727 (#153306)

     Show description for 64518ca
     

     huydhn

     authored and

     pytorchmergebot

     committed
     64518caView commit details

     Copy full SHA for 64518ca

     View code at this pointBrowse repository at this point

   • Revert "refine fp32 precision api (#125888)"

     Show description for fdc387e
     pytorchmergebot

     committed
     fdc387eView commit details

     Copy full SHA for fdc387e

     View code at this pointBrowse repository at this point
3. Commits on May 10, 2025

   • refine fp32 precision api (#125888)

     Show description for 4c11b26
     

     2 peopleauthored and

     pytorchmergebot

     committed
     4c11b26View commit details

     Copy full SHA for 4c11b26

     View code at this pointBrowse repository at this point

   • error out on negative offs or on K=0 in group gemm (#153226)

     Show description for 9c99ea2
     

     ngimel

     authored and

     pytorchmergebot

     committed
     9c99ea2View commit details

     Copy full SHA for 9c99ea2

     View code at this pointBrowse repository at this point
4. Commits on May 9, 2025

   • Move mps_linear forward to use MPS kernels directly instead of MPSGraph (#152210)

     Show description for 4e24ee7
     

     2 peopleauthored and

     pytorchmergebot

     committed
     4e24ee7View commit details

     Copy full SHA for 4e24ee7

     View code at this pointBrowse repository at this point

   • [BE][MPS] Use `squeeze`/`unsqueeze` in Linear (#153288)

     Show description for d07fbd4
     

     malfet

     authored and

     pytorchmergebot

     committed
     d07fbd4View commit details

     Copy full SHA for d07fbd4

     View code at this pointBrowse repository at this point

   • [dynamic shapes] guard_or_false for computeStorageNbytes (#150483)

     Show description for d808a3e
     

     pianpwk

     authored and

     pytorchmergebot

     committed
     d808a3eView commit details

     Copy full SHA for d808a3e

     View code at this pointBrowse repository at this point

   • Format all headers under ATen/cpu/vec, not just top-level (#152364)

     Show description for ec24f8f
     

     swolchok

     authored and

     pytorchmergebot

     committed
     ec24f8fView commit details

     Copy full SHA for ec24f8f

     View code at this pointBrowse repository at this point

   • Add device guard for xpu conv on multi device (#153067)

     Show description for e06a080
     

     guangyey

     authored and

     pytorchmergebot

     committed
     e06a080View commit details

     Copy full SHA for e06a080

     View code at this pointBrowse repository at this point

   • [CUDA][cuBLASLt] Fix scale setting for `allowFP16AccumulationCuBLAS` `true` case (#153083)

     Show description for b30d276
     

     eqy

     authored and

     pytorchmergebot

     committed
     b30d276View commit details

     Copy full SHA for b30d276

     View code at this pointBrowse repository at this point
5. Commits on May 8, 2025

   • [ROCm] Maxpool forward NHWC Perf Improvement targeting Resnet scenarios (#151727)

     Show description for 5683965
     

     amd-hhashemi

     authored and

     pytorchmergebot

     committed
     5683965View commit details

     Copy full SHA for 5683965

     View code at this pointBrowse repository at this point

   • Avoid std::chrono::system_clock (#153135)

     Show description for d291fa8
     

     cyyever

     authored and

     pytorchmergebot

     committed
     d291fa8View commit details

     Copy full SHA for d291fa8

     View code at this pointBrowse repository at this point

   • [ROCm] Upgrade ROCm CI to ROCm6.4 (#151368)

     Show description for fe8ebac
     

     2 peopleauthored and

     pytorchmergebot

     committed
     fe8ebacView commit details

     Copy full SHA for fe8ebac

     View code at this pointBrowse repository at this point

   • Clean up of CUTLASS_VERSION (#152947)

     Show description for dfb91a6
     

     narekmalk

     authored and

     pytorchmergebot

     committed
     dfb91a6View commit details

     Copy full SHA for dfb91a6

     View code at this pointBrowse repository at this point

   • Add torch._C.Tag.needs_contiguous_strides (#152859)

     Show description for 94ca3a4
     

     zou3519

     authored and

     pytorchmergebot

     committed
     94ca3a4View commit details

     Copy full SHA for 94ca3a4

     View code at this pointBrowse repository at this point
6. Commits on May 7, 2025

   • Fix TORCH_CHECK error message in FusedSgdKernel (#153074)

     Show description for 7706074
     

     MeetThePatel

     authored and

     pytorchmergebot

     committed
     7706074View commit details

     Copy full SHA for 7706074

     View code at this pointBrowse repository at this point

   • [CUDA][cuDNN] Fix handling of `CPU` side input and target length tensors in `CTCLoss` (#152745)

     Show description for cecfc7d
     

     eqy

     authored and

     pytorchmergebot

     committed
     cecfc7dView commit details

     Copy full SHA for cecfc7d

     View code at this pointBrowse repository at this point

   • Use gather in index_select (#151715)

     Show description for d042ec8
     

     isuruf

     authored and

     pytorchmergebot

     committed
     d042ec8View commit details

     Copy full SHA for d042ec8

     View code at this pointBrowse repository at this point

   • [MPS] SDPA specialized kernels (#152781)

     Show description for 56492bf
     

     Isalia20

     authored and

     pytorchmergebot

     committed
     56492bfView commit details

     Copy full SHA for 56492bf

     View code at this pointBrowse repository at this point

   • Move additional MPS Unary ops to Iterator (#152876)

     Show description for 327d1b6
     

     Siddharth Kotapati

     authored and

     pytorchmergebot

     committed
     327d1b6View commit details

     Copy full SHA for 327d1b6

     View code at this pointBrowse repository at this point
7. Commits on May 6, 2025

   • [ROCm] opportunistic fastatomics - fix build error with newer compilers (#152841)

     Show description for 8faa0b1
     

     pragupta

     authored and

     pytorchmergebot

     committed
     8faa0b1View commit details

     Copy full SHA for 8faa0b1

     View code at this pointBrowse repository at this point

   • [float16]: Fix the accumulation type for dot and gemv (#152676)

     Show description for 08f5371
     

     f2013519

     authored and

     pytorchmergebot

     committed
     08f5371View commit details

     Copy full SHA for 08f5371

     View code at this pointBrowse repository at this point

   • [float16]: Fast path for torch.dot with float16/bfloat16 (#152799)

     Show description for f47bf38
     

     f2013519

     authored and

     pytorchmergebot

     committed
     f47bf38View commit details

     Copy full SHA for f47bf38

     View code at this pointBrowse repository at this point

   • Revert "Make device check error message more descriptive (#150750)"

     Show description for 451d652
     pytorchmergebot

     committed
     451d652View commit details

     Copy full SHA for 451d652

     View code at this pointBrowse repository at this point

   • Make device check error message more descriptive (#150750)

     Show description for 8253970
     

     zeshengzong

     authored and

     pytorchmergebot

     committed
     8253970View commit details

     Copy full SHA for 8253970

     View code at this pointBrowse repository at this point
8. Commits on May 5, 2025

   • [fbgemm] Implement __obj_flatten__ for LinearPackedParamsBase (#152619)

     Show description for d547c7e
     

     hl475

     authored and

     pytorchmergebot

     committed
     d547c7eView commit details

     Copy full SHA for d547c7e

     View code at this pointBrowse repository at this point

   • Move warning from item to specific number conversions (#152709)

     Show description for 22d1359
     

     albanD

     authored and

     pytorchmergebot

     committed
     22d1359View commit details

     Copy full SHA for 22d1359

     View code at this pointBrowse repository at this point

   • Revert "[float16]: Fast path for torch.dot with float16/bfloat16 (#152799)"

     Show description for fdadda2
     pytorchmergebot

     committed
     fdadda2View commit details

     Copy full SHA for fdadda2

     View code at this pointBrowse repository at this point

   • [float16]: Fast path for torch.dot with float16/bfloat16 (#152799)

     Show description for d57bf53
     

     f2013519

     authored and

     pytorchmergebot

     committed
     d57bf53View commit details

     Copy full SHA for d57bf53

     View code at this pointBrowse repository at this point

   • [dynamo] Guard serialization for FUNCTORCH_STACK_MATCH (#152616)

     Show description for ffd5829
     

     zhxchen17

     authored and

     pytorchmergebot

     committed
     ffd5829View commit details

     Copy full SHA for ffd5829

     View code at this pointBrowse repository at this point

   • Disable SLEEF implementation of vec::maximum in vec128_float_neon.h | Accelerate aten::hardtanh_ by 21x (#152538)

     Show description for 35475a3
     

     Rohanjames1997

     authored and

     pytorchmergebot

     committed
     35475a3View commit details

     Copy full SHA for 35475a3

     View code at this pointBrowse repository at this point

   • Add "#pragma once" to CachingHostAllocator.h (#152800)

     Show description for a1516d9
     

     jhapradip

     authored and

     pytorchmergebot

     committed
     a1516d9View commit details

     Copy full SHA for a1516d9

     View code at this pointBrowse repository at this point

   • [MPS] Migrate div roudning modes (#152758)

     Show description for 0ffd31d
     

     malfet

     authored and

     pytorchmergebot

     committed
     0ffd31dView commit details

     Copy full SHA for 0ffd31d

     View code at this pointBrowse repository at this point