strisunshinewentingwang/pytorch

Pull Requests Commits

[FSDP][BE] Skip asan (#87729)

rohan-varma committed 1 year ago

20c08f29

aten cpu and xnnpack to be compatible with arvr mode build (#87125)

minh2910 committed 1 year ago

bd4c4537

Fix CODE level usage in dynamo config.py (#87522)

williamwen42 committed 1 year ago

a605a307

Added gm.print_readable to torchinductor_trace output (#87717)

Chillee committed 1 year ago

e150a621

[xnnpack][lite-int][graph-build] graph passes and op checking (#87128)

mcr229 committed 1 year ago

b013eb54

Fix debug dir bugs and minifier output directories (#87682)

mlazos committed 1 year ago

44d7ba7e

Intercept aten._reshape_alias for nvFuser (#87072)

IvanYashchuk committed 1 year ago

ff2569bc

Fix typos under functorch directory (#87663)

kiszk committed 1 year ago

a3d495bd

Fix stride for prims.where (#87563)

SherlockNoMad committed 1 year ago

0b162f5b

[Dynamo] Symbolic shape guards (#87570)

voznesenskym committed 1 year ago

bc194948

[ao] Adding FAQ to docs (#87322)

HDCharles committed 1 year ago

d0e12d1c

Fix _refs for aten.zeros/ones/empty/randn (#87569)

SherlockNoMad committed 1 year ago

ece3758a

[inductor] Revert channels-last support (#87588)

anijain2305 committed 1 year ago

ebe5aad4

Fixed minor typos in torch.flip and torch.rot90 (#87724)

scaomath committed 1 year ago

312628d2

[ONNX] Fix pad Circular Mode (#86984)

titaiwangms committed 1 year ago

52ac8adc

Use setup_instance script to enable conda and load cuda libraries (#87296)

xuzhao9 committed 1 year ago

e532fb9a

build: support DNNL_GRAPH_CPU_RUNTIME=TBB (#87512)

min-jean-cho committed 1 year ago

7a6808c5

Add prepend argument to nn.Module hooks (#87370)

mrshenli committed 1 year ago

82698b89

[ONNX] replace AT_ASSERT with TORCH_INTERTNAL_ASSERT take 2 (#86405)

titaiwangms committed 1 year ago

82dff8ee

[ONNX] Support quantized::conv1d_relu (#85997)

titaiwangms committed 1 year ago

65b4a633

Disable test_inductor_timm_shard (#87710)

desertfire committed 1 year ago

15370d32

Graph-break on FSDP in dynamo (#87420)

wconstab committed 1 year ago

874625e0

[pytorch] Layer norm backward speed gain with warp shuffles (#87445)

valentinandrei committed 1 year ago

b6f28334

Add named_buffers to torchdynamo nn_module (#87644)

tugsbayasgalan committed 1 year ago

7b597825

consider numel args when identifying aligned args (#87394)

stumpOS committed 1 year ago

8a2a4ed4

Add get_guard_expr to symbolic_shapes which returns all guards in a single expression (#87665)

Chillee committed 1 year ago

569eebb4

Prefer python meta function over c++ meta function (#87426)

SherlockNoMad committed 1 year ago

eb99c1ef

[ONNX] Add Support on 0d tensor Broadcast (#87211)

titaiwangms committed 1 year ago

65601f5e

Revert "Intercept aten._reshape_alias for nvFuser (#87072)"

pytorchmergebot committed 1 year ago

5308886e

Performance improvment to cumulative seq len (#87530)

drisspg committed 1 year ago

0cba7888

Newer Older