strisunshinewentingwang/pytorch

Pull Requests Commits

Add 64-bit indexing support to THC index reductions (#33405)

peterbell10 committed 4 years ago

c882425c

[caffe2] use Clang identification macro in various places (#33574)

Igor Sugak committed 4 years ago

23846d5a

Add instructions and operators for new bytecode format of PyText model (#33555)

Martin Yuan committed 4 years ago

5782758b

[caffe2] fix invalid % escape in inline assembly strings (#33554)

Igor Sugak committed 4 years ago

108fc783

torch.tensor can infer complex dtype now (#33361)

anjali411 committed 4 years ago

e5cf7afd

Added tensor.is_complex(), is_complex and dtype.is_complex py binding, tensor printing, and dixed the scalar type returned for complex float (#33268)

anjali411 committed 4 years ago

13e4ee78

run peephole to do profile-based optimizations (#33337)

Krovatkin committed 4 years ago

36d724c9

Check for consistent devices in at::where (#33432)

vishwakftw committed 4 years ago

1a257473

Revert D20006312: Revert D19975410: Update documentation on why _cudnn_init_dropout_state looks the way it is.

ezyang committed 4 years ago

71225ecc

Revert D19975411: Remove special case codegen for tril_indices/triu_indices.

VitalyFedyunin committed 4 years ago

687a7e4a

Add missing weight_decay parameter validation for Adam and AdamW (#33126)

jettify committed 4 years ago

d19a50bf

Fix LambdaLR scheduler side effects (#32848)

andfoy committed 4 years ago

cdf381c9

Revert D19975410: Update documentation on why _cudnn_init_dropout_state looks the way it is.

VitalyFedyunin committed 4 years ago

3233033a

Add ability to enable/disable MIOpen at runtime (#33118)

jithunnair-amd committed 4 years ago

718c538f

allow remote torchscript call to itself (#32990)

zhaojuanmao committed 4 years ago

01e1de82

Update documentation on why _cudnn_init_dropout_state looks the way it is. (#33347)

ezyang committed 4 years ago

a9e4448d

Remove special case codegen for tril_indices/triu_indices. (#33305)

ezyang committed 4 years ago

196fda5a

Revert "Disable flaky test TestCppExtensionAOT.test_cuda_extension in… (#33404)

peterjc123 committed 4 years ago

ffe327f7

Revert D19964089: [pytorch][PR] Allow vectorized gpu loop to have different argument types

VitalyFedyunin committed 4 years ago

05fb1600

Delete build_variables.bzl following configerator change.

ezyang committed 4 years ago

883b18ea

[caffe2] make fused rowwise quant/dequant op work for N-dim tensors (#33426)

jspark1105 committed 4 years ago

e95282ab

Updating ONNX checker logic. (#33522)

Spandan Tiwari committed 4 years ago

bf0951d9

Allow vectorized gpu loop to have different argument types (#33222)

zasdfgbnm committed 4 years ago

1fe635be

[Caffe2][ThreadPool] Make sure numThreads does not exceed the number of big cores (#33523)

Hao Lu committed 4 years ago

81394581

[WIP] migrate scatter_ to ATen CPU (+multithreading, nondeterministic) (#33139)

nikitaved committed 4 years ago

602ef0d9

Back out "Revert D19871946: [distributed] pass in timeout to TCP store when initializing" (#33434)

rohan-varma committed 4 years ago

6cb9e6b0

Support broadcast for quantized mul kernel (#30442)

lly-zero-one committed 4 years ago

ecb05f12

Make slow_conv_transpose2d_backward tensors contiguous (#33462)

VitalyFedyunin committed 4 years ago

ea514c81

[caffe2] simplify relative error expr (#32999)

gaurav1086 committed 4 years ago

e5a02aa2

avoid large vector copy when query per_channel q_params (#31040)

zhuhaozhe committed 4 years ago

bd3c6e8e

Newer Older