strisunshinewentingwang/pytorch

Pull Requests Commits

Added handling bf16 hook for older cuda and nccl

Olga Andreeva committed 2 years ago

35c39b7f

Olga Andreeva committed 2 years ago

1b0ad0eb

Olga Andreeva committed 2 years ago

65d26757

Adding 2 lower precision communication hooks for FSDP

Olga Andreeva committed 2 years ago

32c2f691

Added low-precision hooks to fsdp_coom_hooks

Olga Andreeva committed 2 years ago

4652cabb

Use standard mechanism for stdlib names (#81520)

Erotemic committed 2 years ago

4c9eae33

[Py3.10] Allow floats to be imported as Long (#81372)

malfet committed 2 years ago

69d73345

Added note about katex install for building docs (#81550)

Fabio Rocha committed 2 years ago

e100e6fd

[JIT] Implement may_contain_alias in FunctionSchema (#81352)

goldenxuett committed 2 years ago

e71f4e79

disable src mask for transformer and multiheadattention fastpath (#81277)

erichan1 committed 2 years ago

23088fcf

[FSDP] Construct FQN in _full_post_state_dict_hook (#81253)

fegin committed 2 years ago

3ea1b9be

Apply ufmt linter to all py files under test/onnx (#81335)

huydhn committed 2 years ago

88e1c5c1

Back out "Revert D37720837: Back out "Revert D37228314: [Profiler] Include ActivityType from Kineto"" (#81450)

Taylor Robie committed 2 years ago

9d3c35d1

[retake2][mobile] Fix lightweight dispatch OOM error by introducing selective build (#80791)

larryliu0820 committed 2 years ago

e3451385

Fixed the decomposition for `embedding_dense_backward` (#81528)

Chillee committed 2 years ago

5139053e

[ONNX] Add tests to quantized cat (#81484)

justinchuby committed 2 years ago

4c728a75

Stop attempting to install networkx (it's already installed) (#81555)

zou3519 committed 2 years ago

6b82601b

[ONNX] remove outdated ImplicitCastType QA in onnx.rst (#81268)

titaiwangms committed 2 years ago

69608fc5

[MPS] Handle 1D bias for addmm (#81519)

abhudev committed 2 years ago

d68fed56

[vulkan] implement dequantize (#81493)

psavla2 committed 2 years ago

d7d23ffc

[vulkan] implement quantize per tensor (#81492)

psavla2 committed 2 years ago

3a62f9a5

[vulkan] add support for quantized tensors (#81491)

psavla2 committed 2 years ago

7988c531

Make convolution_backward_overrideable give better error (#81538)

ezyang committed 2 years ago

a8f53e80

CPUBlas: Use mkldnn optimized BFloat16 matmul for gemm (#65840)

peterbell10 committed 2 years ago

cc626039

Remove unused variables from state_dict_loader (#81513)

kit1980 committed 2 years ago

d61ae1a7

Remove unused storage_size (#81514)

kit1980 committed 2 years ago

fe34bf12

Remove unused rank from _AllGatherBase backward (#81515)

kit1980 committed 2 years ago

d083b448

[primTorch] Implement fftshift and ifftshift (#80737)

peterbell10 committed 2 years ago

443b13fa

[primTorch] Implement two-dimensional fft transforms (#80736)

peterbell10 committed 2 years ago

c0ff72b3

[primTorch] Implement n-dimensional fft transforms (#80571)

peterbell10 committed 2 years ago

353180e1

Older