strisunshinewentingwang/pytorch

Pull Requests Commits

remove some spurious warnings fixing (#72352)

albanD committed 2 years ago

1bad3c4a

[nnc] Add stride args while running with allocated outputs (#72223)

navahgar committed 2 years ago

ff714299

[FSDP] Add FlatParameter to track the information of a flat parameter (#69241)

fegin committed 2 years ago

224093db

Make LinearPackedParams works with both torchscript and torch.package (#71656)

ShijunK committed 2 years ago

09e2fb8f

[BE] Fix pybind deprecation warnings (#72376)

malfet committed 2 years ago

717d8c62

Fix unused variable warnings (#72410)

r-barnes committed 2 years ago

5da6de5d

Avoid type qualifier specified more than once (#72411)

r-barnes committed 2 years ago

805dff35

Fix unused variable warning (#72412)

r-barnes committed 2 years ago

2b702b43

[PT-D][BE] Fix DDP no_sync() test logic (#72348)

awgu committed 2 years ago

b0479639

Move CUDA linalg code to its own subfolder (#72304)

malfet committed 2 years ago

133461e5

Fix BC by adding aten::_native_multi_head_self_attention (#72429)

janeyx99 committed 2 years ago

d8c3ab11

[PyTorch] Support NVTX range_start and range_end (#70030)

louisfeng committed 2 years ago

83b3b5fb

Tensorimpl cleanup try 2 (#72336)

albanD committed 2 years ago

9f9b9c48

Add ZT fastpath for torch.{dot, vdot} (#71129)

anjali411 committed 2 years ago

9d8f0c78

Update release note bot to actually ping people

albanD committed 2 years ago

4e98a4b6

janeyx99 committed 2 years ago

a004f135

Added prod op to FX2TRT (#72284)

Kelly Stanton committed 2 years ago

9ac28cbf

Define TORCH_ASSERT_ONLY_METHOD_OPERATORS in ATen/core (#72344)

peterbell10 committed 2 years ago

bebf8dd5

dbr quant function fusion [2/x]: use fusion for observation and inference (#71781)

vkuzo committed 2 years ago

998a5adf

fx quant: add fusion matching for operator.add and torch.relu (#71780)

vkuzo committed 2 years ago

d672bbd0

dbr quant function fusion [1/x]: record matches for functions (#71764)

vkuzo committed 2 years ago

5937c48f

dbr quant: record dag of non-quantizeable ops (#71551)

vkuzo committed 2 years ago

7e2a7728

dbr quant: refactor first_call hooks to be standalone (#71324)

vkuzo committed 2 years ago

15808564

dbr quant: rename seen_op_info to seen_q_op_info (#71312)

vkuzo committed 2 years ago

0ff9cf39

[bench] Adding a cpp benchmark to compare performance of nnc with static and symbolic shapes (#72197)

navahgar committed 2 years ago

4eb277ac

[bench] Fix build issues with TensorExpr cpp benchmarks (#72196)

navahgar committed 2 years ago

237e960e

Automated submodule update: FBGEMM (#72158)

facebook-github-bot committed 2 years ago

8559d39c

[Codemod][FBSourceBlackLinter] Daily `arc lint --take BLACK`

CodemodService Bot committed 2 years ago

1edf6f56

Add num_head param to native multihead attention to evade dim_per_head always 64 bug (#72375)

zrphercule2 committed 2 years ago

b9d4c0c7

enables module modification during repackaging (#71520)

PaliC committed 2 years ago

5c6b8975

Newer Older