strisunshinewentingwang/pytorch

Pull Requests Commits

Modularize parsing NCCL_BLOCKING_WAIT in ProcessGroupNCCL (#41076)

osalpekar committed 4 years ago

33f9fbf8

Autograd Doc for Complex Numbers (#41012)

anjali411 committed 4 years ago

db38487e

test nan and inf in TestTorchMathOps (#41225)

ngimel committed 4 years ago

e568b3fa

[caffe2] Add the dedup implementation of fused RowWiseAdagrad op on GPUs (#40282)

jianyuh committed 4 years ago

62e16934

Benchmark RecordFunction overhead on some models (#40952)

Ilia Cherniavskii committed 4 years ago

08227072

Add add_relu fusion pass to optimize_for_mobile. (#40252)

kimishpatel committed 4 years ago

8a79eec9

Added SiLU activation function (#41034)

heitorschueroff committed 4 years ago

75a4862f

Expose private APIs to enable/disable pickling ScriptModules without RPC (#39631)

mrshenli committed 4 years ago

f6eb92a3

[ROCm] Skip caffe2 unique op test for rocm3.5 (#41219)

rohithkrn committed 4 years ago

df252c05

make Int8 FC bias quantization use round flush to infinity

Hector Yuen committed 4 years ago

a79b4168

Revert D22458928: [pytorch][PR] Use explicit templates in CUDALoops kernels

malfet committed 4 years ago

7c2c752e

JIT pass for add relu fusion. (#39343)

kimishpatel committed 4 years ago

c5dcf056

Add fused add_relu op. (#39342)

kimishpatel committed 4 years ago

82c9f79e

[Vec256][neon] Add neon backend for vec256 (#39341)

kimishpatel committed 4 years ago

d6feb614

Add benchmark for add op. (#40059)

kimishpatel committed 4 years ago

bddba1e3

[RPC docs] Remove mention of TensorPipe's SHM and CMA backends as they're not built (#41200)

lw committed 4 years ago

dde3d5f4

restore old documentation references (#39086)

mattip committed 4 years ago

a88099ba

Preserve CUDA gencode flags (#41173)

malfet committed 4 years ago

b952eaf6

Use explicit templates in CUDALoops kernels (#41059)

malfet committed 4 years ago

e3742807

Revert D21870844: Create lazy_dyndeps to avoid caffe2 import costs.

malfet committed 4 years ago

1f135148

add clang code coverage compile flags (#41103)

scintiller committed 4 years ago

22f940b7

Fix max_pool2d perf regression (#41174)

xwang233 committed 4 years ago

2cf31fb5

Make IterableDataset dataloader.__len__ warning clearer (#41175)

ssnl committed 4 years ago

1922f221

[JIT] Fix JIT triage workflow (#41170)

Meghan Lele committed 4 years ago

e84ef45d

[quant][refactor] test_only_eval_fn (#41078)

jerryzh168 committed 4 years ago

c1fa74b2

Don't add NCCL dependency to gloo if system NCCL is used (#41180)

Flamefire committed 4 years ago

7c29a4e6

[caffe2] Fix spatial_batch_norm_op dividision-by-zero crash (#40806)

HC Zhu committed 4 years ago

2252188e

add null check for c2 tensor conversion (#41096)

linbinyu committed 4 years ago

df1f8a48

Print raising warnings in Python rather than C++ if other error occurs (#41116)

t-vi committed 4 years ago

a318234e

Create lazy_dyndeps to avoid caffe2 import costs. (#39488)

c00w committed 4 years ago

07fd5f8f

Newer Older