strisunshinewentingwang/pytorch

Pull Requests Commits

[shard] make state_dict hook be consistent

wanchaol committed 2 years ago

bef2fecb

Add test for torchscripting nn.TransformerEncoder, including fast path (#79796)

erichan1 committed 2 years ago

06274d7a

[AMP] Use generic autocast in example, specify dtype (#79579)

eqy committed 2 years ago

eff74ed7

include cudnn_batch_norm, filter out view

SherlockNoMad committed 2 years ago

64ffc6f8

Create a new Ubunutu 22.04 (jammy) build for platform010 (#77591)

mehtanirav committed 2 years ago

71d82917

filter partitions

SherlockNoMad committed 2 years ago

cf9eab4a

started using mode-based tracing

Chillee committed 2 years ago

f5d7e5a1

[JIT] Add basic aliasing checks for tensor inputs

goldenxuett committed 2 years ago

1432a3d6

include get_item in the fusion group

SherlockNoMad committed 2 years ago

32313161

Add symmetric version of gpu_kernel_with_scalars

peterbell10 committed 2 years ago

e8727994

Add option for allowing non-fake inputs, add deepcopy impl

eellison committed 2 years ago

268bbecf

Fix forward AD add.Scalar to not set tangent as-is

pytorchmergebot committed 2 years ago

e213c6ad

Fix forward AD copy_ into same-sized tensor without fw grad

pytorchmergebot committed 2 years ago

a2d159e6

Remove python implementation for eigh meta

lezcano committed 2 years ago

648a6658

fx quant: refactor qconfig setting out of find_matches

vkuzo committed 2 years ago

7b4e92ac

Support dropout(nested tensor) (#79318)

YifanShenSZ committed 2 years ago

1b25aa67

Use nvfuser's suportted aten ops; Validate partitioner on torchbench

SherlockNoMad committed 2 years ago

75ec3bd1

[DOC] Missing line in serialization notes (#79454)

CompRhys committed 2 years ago

62ba548c

automate hash pinning (#79470)

clee2000 committed 2 years ago

514bf9fe

Revert "Make l1_loss composite"

pytorchmergebot committed 2 years ago

d4a94387

Update build targets to include generated enum_tag.cpp

anjali411 committed 2 years ago

321cd14a

[DataPipe] Enable profiler record context in __next__ branch

NivekT committed 2 years ago

e8ed16f3

[DataPipe] Refactor _hook_iterator for readability

NivekT committed 2 years ago

25ca0067

Make `LR scheduler` stub complete, including `OneCycleLR` and class attributes. (#59476)

sbrodehl committed 2 years ago

fb9d8de3

[NVFuser] add "canBeEnabled" interface

davidberard98 committed 2 years ago

459090e3

reland Add offsets-based reduction to segment_reduce (CPU, CUDA)

mikaylagawarecki committed 2 years ago

7360b53f

Use cuda instead of cudatoolkit for cuda 11.6 (#77164)

atalman committed 2 years ago

64f3742b

[fix] composite compliance: isclose (#79700)

kshitij12345 committed 2 years ago

bcc4dba4

Fix typo in ChainedScheduler docstring (#79775)

bmsknight committed 2 years ago

9acbaaaf

[fix] composite compliance: eig, eigh, symeig (#79698)

kshitij12345 committed 2 years ago

04b98df8

Newer Older