strisunshinewentingwang/pytorch

Pull Requests Commits

torch.jit.script escape hatch (#106229)

pmeier committed 1 year ago

a926be39

Revert "Add initial support for FP8 ONNX export (#106379)"

pytorchmergebot committed 1 year ago

71be8f22

[indcutor] add one triton config for reduction (#106925)

shunting314 committed 1 year ago

e18ca402

[inductor] make thread order consistent with loop order (#106827)

shunting314 committed 1 year ago

6696a75e

Revert "[export] Refactor `constrain_as_value` and `constrain_as_size` (#106591)"

pytorchmergebot committed 1 year ago

745d29b0

Add ONNX export support for huggingface's bigscience/bloom-560m model (#106930)

Thiago Crepaldi committed 1 year ago

0b05aef8

SymInt'ify tile (#106933)

ezyang committed 1 year ago

9f26503b

`asarray`: take the default device into consideration. (#106779)

ysiraichi committed 1 year ago

a5d841ef

Support complex inputs in `nan_to_num` (#106944)

kurtamohler committed 1 year ago

171341ee

[test_nn] add custom device support for dropout tests、lazy_modules te… (#106609)

summerdo committed 1 year ago

7db6eb71

adding mixed_dtype_mm to torch._inductor (#106443)

HDCharles committed 1 year ago

03414081

[export] Refactor `constrain_as_value` and `constrain_as_size` (#106591)

tugsbayasgalan committed 1 year ago

18989890

inductor: fix compile error for inplace variable multi-defined (#106852)

XiaobingSuper committed 1 year ago

df6aaf7b

Causing internal clang tidy to error (#106895)

drisspg committed 1 year ago

7460adf7

[Dynamo x FSDP][1/x] Builder support for deque, appendleft (#106884)

voznesenskym committed 1 year ago

71a336ef

make sure mkldnn convolution given same stride as ref path for nc11 contiguous input (#106966)

XiaobingSuper committed 1 year ago

4df84c3b

NumPy support in torch.compile (#106211)

lezcano committed 1 year ago

a9dca534

[inductor] Use shape env bounds in inductor bounds.py (#106175) (#106568)

masnesral committed 1 year ago

8f774330

[Vulkan] Introduce GPU Memory Layout qualifier (#106978)

SS-JIA committed 1 year ago

62b30180

Add _unsafe_index decomp (#106814)

SS-JIA committed 1 year ago

8c8477e5

[pytorch][ao] Add `torch.matmul` in FloatFunctional/QFunctional (#106831)

jiaxuzhu92 committed 1 year ago

152203d3

[caffe2] Add enforce inside ScatterAssignOp (#106882)

chowarfb committed 1 year ago

dfb1b959

Reword release version: major->minor in README.md (#106980)

atalman committed 1 year ago

aef27bdb

[inductor] Enable multilayer reductions with dynamic shapes (#106747)

peterbell10 committed 1 year ago

a62de2d5

[inductor] Type triton size arguments in the kernel index_dtype (#106870)

peterbell10 committed 1 year ago

fa65df37

[inductor] Fix reference_as_float gradcheck (#106626)

peterbell10 committed 1 year ago

3b2cb459

Fix some typos, mostly "that that" (#106901)

apivovarov committed 1 year ago

02abbb81

[FSDP] Fix train -> EMA -> eval with mixed precision (#106858)

awgu committed 1 year ago

7b94d934

[quant][pt2e][fix] Remove the requirement of using no_grad for reference model that contains quantized conv2d (#106924)

jerryzh168 committed 1 year ago

79449e62

Make RNGStateTracker support cuda-like device (#106771)

alanhe151220037 committed 1 year ago

1afbc985

Newer Older