strisunshinewentingwang/pytorch

Pull Requests Commits

Add experimental torch.export prototype (#95070)

tugsbayasgalan committed 1 year ago

454c48b9

Revert "Use FindCUDAToolkit to find cuda dependencies (#82695)"

pytorchmergebot committed 1 year ago

801b3f8f

Node.stack_trace should have innermost frame last (#95592)

SherlockNoMad committed 1 year ago

f8692dcc

better error message when functionalization cant handle op (#95392)

bdhirsh committed 1 year ago

b818b3fe

fix embedding_backward_dense decomp with broadcasting (#95499)

bdhirsh committed 1 year ago

ddd6b53d

fix primtorch handling for sub.scalar with alpha and float64 arg (#95421)

bdhirsh committed 1 year ago

84e2d957

Better mark_dynamic assertions (#95566)

voznesenskym committed 1 year ago

eff5ae87

Add super().setUp() in test_symbolic_shape_analysis (#95336)

clee2000 committed 1 year ago

4e926db1

Update copyright (#95652)

Svetlana Karslioglu committed 1 year ago

d7146e78

[jit] Add shapes info to the output type of CallFunction nodes after tracing, if the output is a tensor (#95544)

davidberard98 committed 1 year ago

10bf019b

Remove mentions of distributed/_shard/test_replicated_tensor (#95632)

clee2000 committed 1 year ago

5272d6e6

[4/N][Deprecate ST][BE] Move warnings of Partial Tensor to functions (#95631)

fduwjj committed 1 year ago

38fdd28d

Revert "Convert operator.not_ to torch.logical_not (#94626)"

pytorchmergebot committed 1 year ago

33cf6235

Inductor allgather_into_tensor (#95530)

wconstab committed 1 year ago

cc6da7b9

Support elementwise add / mul for [B, *] nested, [B, 1] dense (CUDA only) (#95620)

jbschlosser committed 1 year ago

68eec90c

Add _int_mm to expose cuBLAS int8@int8 -> int32 matmul (#94339)

cpuhrsch committed 1 year ago

1fe2a9d1

make overriding operator warning message only print once (#95179)

heidongxianhua committed 1 year ago

32558910

[NCCL] (re-open) Optionally avoid `recordStream` calls in `ProcessGroupNCCL` (#89880)

eqy committed 1 year ago

29f9a702

[meta_tensor] polish error strings in meta registrations (#95052)

1SAA committed 1 year ago

f43ce955

[CI] Do not compare two eager run results against fp64 result (#95616)

desertfire committed 1 year ago

fa5a4b0d

dynamo export should be able to export identity function (#94962)

qihqi committed 1 year ago

34617d7e

Re-enable a FX-to-ONNX kwargs Test (#94763)

wschin committed 1 year ago

868640e0

Update `fx.pass.graph_drawer` usage doc to draw fx graph (#95534)

wenzhe-nrv committed 1 year ago

8dfac7b8

Cleanup Windows warning suppression in CMake and fix some warnings in the source code (#94927)

cyyever committed 1 year ago

f27e09de

Revert "[Functional Collectives] Migrate DeviceMesh::all_reduce to use functional all_reduce. (#95009)"

pytorchmergebot committed 1 year ago

d950f455

Add bfloat16 support to upsample (#95500)

jamaliki committed 1 year ago

1cf11c1c

Fix split_module bug (#95493)

RenfeiChen-FB committed 1 year ago

c44a7330

[Quant] Fix setting fixed qparams for inner LSTM ops (#95537)

andrewor14 committed 1 year ago

a3b505c5

Fix typos in documents under torch (#95597)

kiszk committed 1 year ago

31ce32b0

[dynamo] Fix keyword argument name of all_dim (#95600)

kiszk committed 1 year ago

3beb6445

Newer Older