Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 28, 2025

• [DTensor] Error on illegal view op during sharding prop (#149764)

  Show description for 0c52ee1
  

  wconstab

  authored and

  pytorchmergebot

  committed
  0c52ee1

  Copy full SHA for 0c52ee1

• [DTensor] make test_dtensor_ops report dtensor_args (#152045)

  Show description for efeed72
  

  wconstab

  authored and

  pytorchmergebot

  committed
  efeed72

  Copy full SHA for efeed72

Commits on Apr 27, 2025

• Fix broken URLs (#152237)

  Show description for e2f9759
  

  shoumikhin

  authored and

  pytorchmergebot

  committed
  e2f9759

  Copy full SHA for e2f9759

Commits on Apr 25, 2025

• [CP] Use TorchFunctionMode to dispatch SDPA for CP (#147902)

  Show description for 6aa9280
  

  fegin

  authored and

  pytorchmergebot

  committed
  6aa9280

  Copy full SHA for 6aa9280

Commits on Apr 16, 2025

• improve noop elimination for view (#151095)

  Show description for 2fd2692
  

  BoyuanFeng

  authored and

  pytorchmergebot

  committed
  2fd2692

  Copy full SHA for 2fd2692

Commits on Apr 14, 2025

• [ROCm][CI] Enable distributed CI on MI300 (#150667)

  Show description for 001695c
  

  jithunnair-amd

  authored and

  pytorchmergebot

  committed
  001695c

  Copy full SHA for 001695c

Commits on Apr 13, 2025

• [DTensor] Add DTensor redistribute fwd/bwd datatype conversion to enable SimpleFSDP mixed precision training (#150740)

  Show description for 1c5619e
  

  ruisizhang123

  authored and

  pytorchmergebot

  committed
  1c5619e

  Copy full SHA for 1c5619e

Commits on Apr 12, 2025

• [dtensor] add op support for torch._grouped_mm (#151072)

  Show description for 7dd2ed1
  

  tianyu-l

  authored and

  pytorchmergebot

  committed
  7dd2ed1

  Copy full SHA for 7dd2ed1

Commits on Apr 11, 2025

• [DTensor] Fix empty shard global-offset calculation (#150862)

  Show description for c3bc6b3
  

  wconstab

  authored and

  pytorchmergebot

  committed
  c3bc6b3

  Copy full SHA for c3bc6b3

• [dtensor] add op support for torch.cumsum (#151071)

  Show description for d385179
  

  tianyu-l

  authored and

  pytorchmergebot

  committed
  d385179

  Copy full SHA for d385179

Commits on Apr 10, 2025

• [async TP] Fix handling of case where scatter dim = 0 for 2D output tensor (#150935)

  Show description for e7ed50f
  

  danielvegamyhre

  authored and

  pytorchmergebot

  committed
  e7ed50f

  Copy full SHA for e7ed50f

Commits on Apr 9, 2025

• [DTensor] clean up _local_shard_size_and_offset (#150650)

  Show description for a8b48ff
  

  wconstab

  authored and

  pytorchmergebot

  committed
  a8b48ff

  Copy full SHA for a8b48ff

• [DTensor] StridedShard support uneven sharding (#150490)

  Show description for 3532dd4
  

  wconstab

  authored and

  pytorchmergebot

  committed
  3532dd4

  Copy full SHA for 3532dd4

• [DTensor] add _explicit_order_placements util (#150493)

  Show description for c59aaa0
  

  wconstab

  authored and

  pytorchmergebot

  committed
  c59aaa0

  Copy full SHA for c59aaa0

Commits on Apr 3, 2025

• [export] specialize for aten.to (#149235)

  Show description for 90ddb33
  

  pianpwk

  authored and

  pytorchmergebot

  committed
  90ddb33

  Copy full SHA for 90ddb33

Commits on Apr 1, 2025

• [dtensor][tp] add a ParallelStyle PrepareModuleInputOutput (#150372)

  Show description for d2ad9aa
  

  tianyu-l

  authored and

  pytorchmergebot

  committed
  d2ad9aa

  Copy full SHA for d2ad9aa

• [dtensor] add op support for select_backward and slice_backward (#150357)

  Show description for 5d6ac2d
  

  tianyu-l

  authored and

  pytorchmergebot

  committed
  5d6ac2d

  Copy full SHA for 5d6ac2d

Commits on Mar 31, 2025

• [ROCm] Enable several fsdp related UTs (#149369)

  Show description for 47cdad2
  

  pragupta

  authored and

  pytorchmergebot

  committed
  47cdad2

  Copy full SHA for 47cdad2

Commits on Mar 30, 2025

• [Async TP] Fuse matmul-reduce-scatters when reduce scatters have multiple users, and save fused node for backward instead of reduce_scatter node (#149946)

  Show description for 157bff2
  

  danielvegamyhre

  authored and

  pytorchmergebot

  committed
  157bff2

  Copy full SHA for 157bff2

Commits on Mar 27, 2025

• Implement aten.select.int sharding strategy (#149842)

  Show description for 6841451
  

  kkkkeeee

  authored and

  pytorchmergebot

  committed
  6841451

  Copy full SHA for 6841451

• [Async TP] More robust support for rowwise scales when fusing matmul reduce-scatter (#149247)

  Show description for ae29f05
  

  danielvegamyhre

  authored and

  pytorchmergebot

  committed
  ae29f05

  Copy full SHA for ae29f05

Commits on Mar 16, 2025

• Support subclass constructor capturing in export (#147014)

  Show description for 6b1b95a
  

  tugsbayasgalan

  authored and

  pytorchmergebot

  committed
  6b1b95a

  Copy full SHA for 6b1b95a

Commits on Mar 6, 2025

• [dtensor] add CuDNN SDPA op support to DTensor (#148537)

  Show description for e2a0296
  

  XilunWu

  authored and

  pytorchmergebot

  committed
  e2a0296

  Copy full SHA for e2a0296

Commits on Mar 5, 2025

• Revert "[dtensor] add aten._scaled_dot_product_cudnn_attention.default op support (#148377)"

  Show description for c9edd37
  pytorchmergebot

  committed
  c9edd37

  Copy full SHA for c9edd37

• [dtensor] add aten._scaled_dot_product_cudnn_attention.default op support (#148377)

  Show description for 9eef457
  

  XilunWu

  authored and

  pytorchmergebot

  committed
  9eef457

  Copy full SHA for 9eef457

Commits on Mar 4, 2025

• [dynamo] Support passing arguments to `DeviceMesh.get_group` (#147741)

  Show description for d1abde1
  

  danthe3rd

  authored and

  pytorchmergebot

  committed
  d1abde1

  Copy full SHA for d1abde1

• [dtensor] refactor sharding prop to handle cross mesh computation (#147869)

  Show description for f859722
  

  wanchaol

  authored and

  pytorchmergebot

  committed
  f859722

  Copy full SHA for f859722

Commits on Mar 2, 2025

• [async TP] insert reshape node to handle "reshape -> scaled mm -> reshape pattern" in async TP with rowwise scales (#148001)

  Show description for de7af81
  

  danielvegamyhre

  authored and

  pytorchmergebot

  committed
  de7af81

  Copy full SHA for de7af81

Commits on Mar 1, 2025

• [DTensor][Test] Add a test to demonstrate current dtensor view behavior if redistribution happens (#148015)

  Show description for 2f1b8e0
  

  wz337

  authored and

  pytorchmergebot

  committed
  2f1b8e0

  Copy full SHA for 2f1b8e0

• Revert "[async TP] insert reshape node to handle "reshape -> scaled mm -> reshape pattern" in async TP with rowwise scales (#148001)"

  Show description for 191c9bd
  pytorchmergebot

  committed
  191c9bd

  Copy full SHA for 191c9bd

• [async TP] insert reshape node to handle "reshape -> scaled mm -> reshape pattern" in async TP with rowwise scales (#148001)

  Show description for b8efebe
  

  danielvegamyhre

  authored and

  pytorchmergebot

  committed
  b8efebe

  Copy full SHA for b8efebe

Commits on Feb 28, 2025

• Revert "[async TP] insert reshape node to handle "reshape -> scaled mm -> reshape pattern" in async TP with rowwise scales (#148001)"

  Show description for ebc3f27
  pytorchmergebot

  committed
  ebc3f27

  Copy full SHA for ebc3f27

• [async TP] insert reshape node to handle "reshape -> scaled mm -> reshape pattern" in async TP with rowwise scales (#148001)

  Show description for 6e037ac
  

  danielvegamyhre

  authored and

  pytorchmergebot

  committed
  6e037ac

  Copy full SHA for 6e037ac

• [dtensor][fix] fix _scaled_dot_product_flash_attention sharding (#148125)

  Show description for 4106aa3
  

  XilunWu

  authored and

  pytorchmergebot

  committed
  4106aa3

  Copy full SHA for 4106aa3

Commits on Feb 26, 2025

• [DTensor][random] defer DTensor RNG state sync until first random op call or manual_seed call; support more flexible OffsetBasedRNGTracker init (#147025)

  Show description for ef61c29
  

  XilunWu

  authored and

  pytorchmergebot

  committed
  ef61c29

  Copy full SHA for ef61c29

Pagination

Previous

Next