strisunshinewentingwang/pytorch

Pull Requests Commits

[NCCL] Add Environment Variable to guard Async Error Handling feature (#44163)

osalpekar committed 4 years ago

48c47db8

[NCCL] ProcessGroupNCCL Destructor Blocks on WorkNCCL Completion (#41054)

osalpekar committed 4 years ago

211ece72

[NCCL] WorkNCCL Helper Functions (#41053)

osalpekar committed 4 years ago

afbf2f14

[NCCL] Abort Errored and Timed Out NCCL Communicators from Watchdog Thread (#41052)

osalpekar committed 4 years ago

f8f7b784

[NCCL] Use cudaEventQuery to Poll for GPU operation errors (#41051)

osalpekar committed 4 years ago

4e5c55ef

[NCCL] Timeout Loop Thread for Async Error Handling (#41050)

osalpekar committed 4 years ago

1df24fd4

Preserve .ninja_log in build artifacts (#44390)

malfet committed 4 years ago

15cbd1cf

[Reland] Optimize code path for adaptive_avg_pool2d when output size is (1, 1) (#44211)

xwang233 committed 4 years ago

ef4475f9

Benchmarks: make fuser and executor configurable from command line. (#44291)

Mikhail Zolotukhin committed 4 years ago

37093f4d

Misc. FakeLowP OSS cleanup (#44331)

venkatacrc committed 4 years ago

364d03a6

BUG: make cholesky_solve_out do broadcast, error checking (#43137)

mattip committed 4 years ago

758c2b96

Use compile time cudnn version if linking with it statically (#44402)

malfet committed 4 years ago

683380fc

Fix frac in CUDA fuser (#44152)

bertmaher committed 4 years ago

6ec8fabc

Prevent the TE fuser from getting datatypes it can't handle (#44160)

bertmaher committed 4 years ago

350130a6

[te] Fix casting of unsigned char, and abs(int) (#44157)

bertmaher committed 4 years ago

960c088a

Skipping CUDA tests in ProcessGroupGloo and logs (#42488)

osalpekar committed 4 years ago

7c464eed

Autocast wrappers for RNN cell apis (#44296)

mcarilli committed 4 years ago

2a87742f

Moves some of TestTorchMathOps to OpInfos (#44277)

Mike Ruberry committed 4 years ago

a953a825

Disable a test (#44348)

Krovatkin committed 4 years ago

f044b17a

Don't use VCOMP if Intel OMP is used (#44280)

peterjc123 committed 4 years ago

cfd3620b

Remove pybind11 from required submodules (#44278)

Flamefire committed 4 years ago

d23f3170

Dump optimized graph when logging in already-optimized PE (#44315)

bertmaher committed 4 years ago

8acce550

Export Node::isBefore/isAfter for PythonAPI (#44162)

taewookoh committed 4 years ago

7a64b0c2

[Caffe2] Add RMSNormOp (#44338)

xiaomengy committed 4 years ago

135ebbde

Rename test_distributed to test_distributed_fork (#42932)

rohan-varma committed 4 years ago

106459ac

Enable test_distributed to work with spawn mode (#41769)

rohan-varma committed 4 years ago

b22abbe3

[quant] fill_ path for quantized tensors (#43303)

z-a-f committed 4 years ago

1d01fcdc

Resolve Autograd key for disable_variable_dispatch flag. (#44268)

Ailing Zhang committed 4 years ago

4aacfab2

Port nonzero cuda from THC to ATen (#44259)

ngimel committed 4 years ago

ecc6358d

[TensorExpr] Fuser: check node inputs' device before merging the node into a fusion group. (#44241)

Mikhail Zolotukhin committed 4 years ago

bd8e38cd

Newer Older