Commits

Branch selector

 master

User selector

bdhirsh

Datepicker

All time

Commit History

Commits on Aug 6, 2024

• properly register conjugate/neg fallthroughs to prim ops (pytorch#132699)

  Show description for 8d2c272
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  8d2c272

  Copy full SHA for 8d2c272

• torch.autograd.graph.increment_version: accept List[Tensor], use in AOTDispatcher (pytorch#132652)

  Show description for e3394e5
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  e3394e5

  Copy full SHA for e3394e5

• dynamo: use equality guards instead of id guards for Placement/DeviceMesh (pytorch#124401)

  Show description for e6eee04
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  e6eee04

  Copy full SHA for e6eee04

• return_and_correct_aliasing: skip dispatcher when swapping storage (pytorch#132524)

  Show description for 26c6786
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  26c6786

  Copy full SHA for 26c6786

Commits on Aug 5, 2024

• fsdp.set_: convey to functionalization that it mutates storage (pytorch#132322)

  Show description for af8b8a4
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  af8b8a4

  Copy full SHA for af8b8a4

• move torch._functionalize APIs to pybind. add one for marking storage mutations (pytorch#132337)

  Show description for 1a0db29
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  1a0db29

  Copy full SHA for 1a0db29

• make functorch CSE respect mutations as barriers (like fsdp.set_) (pytorch#132243)

  Show description for 4db368a
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  4db368a

  Copy full SHA for 4db368a

• DTensor: add more foreach ops to supported sharding prop list (pytorch#132066)

  Show description for b465a58
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  b465a58

  Copy full SHA for b465a58

Commits on Aug 1, 2024

• fastpath FunctionalTensor sizes() (pytorch#132084)

  Show description for 997f64a
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  997f64a

  Copy full SHA for 997f64a

Commits on Jul 29, 2024

• make _inductor.config.rocm.supported_arch set order deterministic for caching (pytorch#131921)

  Show description for 6cbad37
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  6cbad37

  Copy full SHA for 6cbad37

Commits on Jul 26, 2024

• fast-path FakeTensor detach (pytorch#131899)

  Show description for 071ac38
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  071ac38

  Copy full SHA for 071ac38

• _get_operation_overload: dont raise exception when overload does not exist (pytorch#131554)

  Show description for 5612408
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  5612408

  Copy full SHA for 5612408

• AOTDispatcher: properly bump version counter on input mutations in inference graphs (pytorch#131665)

  Show description for e4ace1a
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  e4ace1a

  Copy full SHA for e4ace1a

• dont dispatch aten.conj(scalar_tensor) back to python (pytorch#131482)

  Show description for 5570a0d
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  5570a0d

  Copy full SHA for 5570a0d

• dynamo: mutations on .data should be invisible to autograd (pytorch#131403)

  Show description for 8bb9aa9
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  8bb9aa9

  Copy full SHA for 8bb9aa9

Commits on Jul 25, 2024

• add python binding for _get_current_graph_task_keep_graph (pytorch#131038)

  Show description for e944386
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  e944386

  Copy full SHA for e944386

Commits on Jul 23, 2024

• avoid OpOverloadPacket.__getattr__ calls in inductor lowering (pytorch#131348)

  Show description for b2ad16f
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  b2ad16f

  Copy full SHA for b2ad16f

Commits on Jul 16, 2024

• inductor: avoiding moving constructor to cuda when it would cause h2d sync in index_put_ fallback (pytorch#130338)

  Show description for 39eeaac
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  39eeaac

  Copy full SHA for 39eeaac

Commits on Jul 3, 2024

• AOTI: dont treat views of buffers as constants (pytorch#129688)

  Show description for 9b902b3
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  9b902b3

  Copy full SHA for 9b902b3

Commits on Jun 27, 2024

• [Traceable FSDP2] Add auto-functionalize support for mutable list[Tensor] (copy from Brian's PR pytorch#127347); enable E2E inductor unit test for transformer model (pytorch#129502)

  Show description for a4d7aa4
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  a4d7aa4

  Copy full SHA for a4d7aa4

Commits on Jun 23, 2024

• [Brian's PR pytorch#128754] Use torch.ops.fsdp.set_ for FSDP2 storage resize; dont functionalize resize_, set_, split_with_sizes_copy.out (pytorch#129203)

  Show description for b91a9dc
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  b91a9dc

  Copy full SHA for b91a9dc

Commits on Jun 21, 2024

• [Brian's PR pytorch#128981] fix dynamo isinstance inlining for nn.Parameter + subclasses (pytorch#129162)

  Show description for 880e894
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  880e894

  Copy full SHA for 880e894

Commits on Jun 20, 2024

• [Traceable FSDP2] [Dynamo] Add tracing support for out-variant custom ops that return None (pytorch#129078)

  Show description for 8c25426
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  8c25426

  Copy full SHA for 8c25426

Commits on Jun 15, 2024

• inductor: pre-grad bmm pass shouldn't match if output is mutated (pytorch#128570)

  Show description for 271852a
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  271852a

  Copy full SHA for 271852a

• FunctionalTensor: dispatch metadata directly to inner tensor (pytorch#127927)

  Show description for ba19ed9
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  ba19ed9

  Copy full SHA for ba19ed9

Commits on Jun 7, 2024

• fix torch.compile with triton kernels under inference_mode (pytorch#124489)

  Show description for 476bfe6
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  476bfe6

  Copy full SHA for 476bfe6

Commits on May 31, 2024

• improve eager overhead of _disable_dynamo (pytorch#127325)

  Show description for 8af1c65
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  8af1c65

  Copy full SHA for 8af1c65

Commits on May 10, 2024

• functionalize storage resizing, minimal ppFSDP traceable forward (pytorch#122434)

  Show description for f25c7c9
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  f25c7c9

  Copy full SHA for f25c7c9

Commits on May 8, 2024

• AsyncCollectiveTensor: prevent wait_tensor() calls on graph inputs from getting DCEd (pytorch#125677)

  Show description for e28d994
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  e28d994

  Copy full SHA for e28d994

• AOTAutograd: use info not debug logging for ViewAndMutationMeta (pytorch#125676)

  Show description for 5d97c22
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  5d97c22

  Copy full SHA for 5d97c22

Commits on May 1, 2024

• Reland "make sure dynamo doesn't inline DTensor __new__ or __torch_dispatch__ (pytorch#123347)" (pytorch#125288)

  Show description for 599a2e2
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  599a2e2

  Copy full SHA for 599a2e2

• AOTAutograd: force tangents to be contiguous when subclass inner tensor is noncontiguous (pytorch#124400)

  Show description for 9e9ba61
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  9e9ba61

  Copy full SHA for 9e9ba61

• fix FakeTensor creation on noncontiguous subclasses (pytorch#124399)

  Show description for 5173cbe
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  5173cbe

  Copy full SHA for 5173cbe

• support as_python_constant on PlacementClassVariable (pytorch#124398)

  Show description for 7058563
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  7058563

  Copy full SHA for 7058563

Commits on Apr 26, 2024

• Test reland "AOTAutograd: gate view-replay behind config, not the def… (pytorch#124948)

  Show description for fc2aa23
  

  bdhirsh

  authored and

  pytorchmergebot

  committed
  fc2aa23

  Copy full SHA for fc2aa23

Pagination

Previous

Next