strisunshinewentingwang/pytorch

Pull Requests Commits

[pytorch] use cublas lt interface for bias fusion (#72148)

jspark1105 committed 2 years ago

3aecce70

add assert to make sure expected number of LTC roots matches what TS … (#73112)

Krovatkin committed 2 years ago

237574db

Adding details to kl.py (#72845)

patel-zeel committed 2 years ago

c837caf5

Documenting cuda 11.5 windows issue (#73013)

atalman committed 2 years ago

46f9e16a

[Easy] Update the bytecode version comment (#73097)

cccclai committed 2 years ago

d059c082

[vulkan] Allow benchmark binary to handle non-single tensor inputs/outputs for Vulkan models (#73109)

SS-JIA committed 2 years ago

52175307

[vulkan] Re-route arithmetic ops to scalar versions when second arg is zero-dim (#73108)

SS-JIA committed 2 years ago

bdc8b3f3

[vulkan] Clamp tanh activation op input to preserve numerical stability (#73107)

SS-JIA committed 2 years ago

564f9922

add BFloat16 sparse operators on CPU: copy, coalesce, sparse_mask, ad… (#72846)

jiayisunx committed 2 years ago

c6f56599

Add simple correctness check for native MHA (#72941)

zrphercule2 committed 2 years ago

f41db99a

[Vulkan] Add performance test for GRU operator (#73126)

beback4u committed 2 years ago

b1bd2268

Fix backward compat (#73127)

malfet committed 2 years ago

534d5cf9

[codemod][type-comments] Convert type comments in api.py (#73084)

stroxler committed 2 years ago

906d26fb

[ONNX] Add pixel_unshuffle support in opset 9

shubhambhokare1 committed 2 years ago

671c8a45

[jit] Add tests for JIT with dynamic shape fusion (#72201)

navahgar committed 2 years ago

0d667489

[codemod][type-comments] Convert type comments in workspace_test.py (#73086)

stroxler committed 2 years ago

374de336

Move native MHA code out of PyTorch core (#72944)

swolchok committed 2 years ago

79a216ce

Use irange in PyTorch (#72836)

sjoakley committed 2 years ago

1646a003

Remove call into python API without GIL being held in c10d (#72928)

albanD committed 2 years ago

7fe3f334

.github: Consolidate binary checkout logic

seemethere committed 2 years ago

9c8fb2ee

[onnx export] Add broadcast to matmul shape inference (#70534)

BowenBao committed 2 years ago

956bafef

[ONNX] Fix an assertion failure involving Slice (#71965)

BowenBao committed 2 years ago

98f9ff90

Integrate full ONNX check into ONNX export API (#71125)

BowenBao committed 2 years ago

2791725a

[Static Runtime] Do not replace with copy variants if TE fuser is enabled (#72946)

navahgar committed 2 years ago

2724e4c0

[Static Runtime] Handle fallback graphs that are generated as part of the TE Fuser (#72945)

navahgar committed 2 years ago

02afdd54

Move magma utils to its own header (#73058)

malfet committed 2 years ago

87f882b0

[ONNX] Add export support for linalg norm (#66575)

BowenBao committed 2 years ago

5843fea9

[ONNX] First version of quantized model export: Support quantized.Linear (#69232)

BowenBao committed 2 years ago

32f6a1e2

[ONNX] Improve Expand shape inference (#69264)

BowenBao committed 2 years ago

a6517c20

Disable test history as it's fragile

janeyx99 committed 2 years ago

08510ba5

Newer Older