strisunshinewentingwang/pytorch

Pull Requests Commits

Adopt upstreamed ltc_ops.h

desertfire committed 2 years ago

2de9902f

Adopt upstreamed tensor_util.h

desertfire committed 2 years ago

3e5fd4e4

Adopt upstreamed ir_dump_util.h

desertfire committed 2 years ago

c443e95f

Adopt upstreamed helpers.h

desertfire committed 2 years ago

ce0f3ef5

Adopt upstreamed LazyView and view ops IR

desertfire committed 2 years ago

41a4c16e

[quant][graphmode][fx] Add default_replay_qconfig for ops like reshape (#69249)

jerryzh168 committed 2 years ago

ca945d98

[JIT] Separate GPU implementation of frozen_conv_add_relu_fusion.cpp (#68149)

davidberard98 committed 2 years ago

8b1e4963

Enable build-split for all CUDA-11.x version (#69494)

malfet committed 2 years ago

e55b9397

[wip][quant][graphmode] produce reference pattern for binary ops and then rewrite to quantized op (#68229)

jerryzh168 committed 2 years ago

bd8a4a93

[fx2trt][easy] add sparse flag to TRTInterpreter (#69495)

842974287 committed 2 years ago

bcd03038

[fx2trt] Separate sign from `trunc_div` and use it for acc_ops.sign (#69486)

842974287 committed 2 years ago

32115883

[fx2trt] [Prep for release] Add type hints to converters and separate main files (#69458)

842974287 committed 2 years ago

e23827e6

[fx2trt] Add a helper function to generate specs for dynamic batch size (#69405)

842974287 committed 2 years ago

a2d1cadf

[fx2trt] Use weights shape as normalize shape in layer norm (#69401)

842974287 committed 2 years ago

cfe3cbb3

Revert D32704467: [Autograd/Checkpoint] Checkpoint implementation without reentrant autograd

suo committed 2 years ago

59e98b66

Initialize upgrader and operator version files (#68772)

tugsbayasgalan committed 2 years ago

bc895289

Improve efficiency of shape hash by not using tostring (#69404)

wconstab committed 2 years ago

Verified 767ce365

[Pytorch Edge] Add new_empty_strided to tracer (#69492)

JacobSzwejbka committed 2 years ago

9e678446

[PyTorch][Distributed] Use auto-grad enabled collections for the shared linear op to enable backward grad calculation (#68096)

fduwjj committed 2 years ago

65b0f389

Deprecate CreateFrom method in LazyTensor

desertfire committed 2 years ago

e532862d

[PyTorch][Distributed] Enable Reduce Scatter and modify all_to_all for sharded linear with more test cases. (#68786)

fduwjj committed 2 years ago

7c2489bd

[Autograd/Checkpoint] Checkpoint implementation without reentrant autograd (#69027)

rohan-varma committed 2 years ago

e032dae3

add VSX dispatch for fft_fill_with_conjugate_symmetry_stub (#68914)

cdeepali committed 2 years ago

4d81175a

.github: Volume mount local netrc for docs push (#69472)

seemethere committed 2 years ago

f87faf3c

[FSDP] Enforce wrapper_cls as a mandatory kwarg in enable_wrap. (#69358)

rohan-varma committed 2 years ago

1859e5f0

[FSDP] Kill config_auto_wrap_policy, remove policy from enable_wrap, (#69357)

rohan-varma committed 2 years ago

00245fed

[FSDP] Remove auto_wrap() (#69356)

rohan-varma committed 2 years ago

c95277e9

[FSDP] Make recursive_wrap, wrap APIs independent of ConfigAutoWrap. (#68776)

rohan-varma committed 2 years ago

f333cde1

FX pass: fuse_sparse_matmul_add (#69340)

Marat Subkhankulov committed 2 years ago

456139d0

[Vulkan] Implement slice operator (#69382)

beback4u committed 2 years ago

68b5c86e

Newer Older