strisunshinewentingwang/pytorch

Pull Requests Commits

tweak thresholds

eqy committed 2 years ago

e807f532

abandon attempt to rely on @tf32_on_and_off

eqy committed 2 years ago

b4c7af5c

eqy committed 2 years ago

b99bcadc

[pyfunctorch] Generate a more meaningful name for _SingleLevelAutogradFunction (#90418)

zou3519 committed 2 years ago

f21cb7d7

Fix circular import in torch/autograd/function.py (#90415)

zou3519 committed 2 years ago

da42eab4

functorch.jvp support for autograd.Function (#90077)

zou3519 committed 2 years ago

4809e838

Run inductor benchmark test for every PR (#90773)

desertfire committed 2 years ago

dcb73aa2

Inductor cpp wrapper: support more dtypes of input (#88666)

chunyuan-w committed 2 years ago

cc4131a8

Move _test_inductor_realize into python (#90517)

peterbell10 committed 2 years ago

ba77afbc

Inductor cpp wrapper: support Reduction (#88561)

chunyuan-w committed 2 years ago

d35aa2f6

symbolic-shapes: -anjali411, +jbschlosser (#90816)

ezyang committed 2 years ago

7963dbf3

Fakify params and weights under private config (#90417)

tugsbayasgalan committed 2 years ago

1aab7553

to_sparse_XXX: backward support (#90281)

nikitaved committed 2 years ago

3870a9e2

Optimized vertical flip using memcpy (#89414)

vfdev-5 committed 2 years ago

708108a9

Fix non-existing parameters in docstrings in torch/onnx (#90593)

kit1980 committed 2 years ago

e54c6c28

inductor: using pre-existing fake mode to fallback kernels (#90814)

XiaobingSuper committed 2 years ago

37cd96a6

Add tracing context, Integrate dynamo guards into torch._guards (#90647)

voznesenskym committed 2 years ago

6c8ef6a4

Fix gradcheck for BSR and BSC inputs. (#90719)

pearu committed 2 years ago

f4099af1

Support (non-batch) BSR/BSC to COO sparse tensor conversions (#90718)

pearu committed 2 years ago

a60d7120

Restore test_warn_types (#90810)

ezyang committed 2 years ago

cc504ce2

Upgrade CI to ROCm5.3 (#88297)

jithunnair-amd committed 2 years ago

e8e591b7

[ao][fx] fixing public v private for pattern_utils.py (#88397)

HDCharles committed 2 years ago

258860fa

[vision hash update] update the pinned vision hash (#90727)

pytorchmergebot committed 2 years ago

76939217

Include dispatch key in wrapper symbol name (#90674)

ezyang committed 2 years ago

e8737013

[FSDP] Skip param check for pure FP16 (#90785)

awgu committed 2 years ago

6c605e9c

Inductor cpp wrapper: support None as output (#88560)

chunyuan-w committed 2 years ago

e2e4a80c

[FSDP][Easy] ufmt files (#90548)

awgu committed 2 years ago

93aee0cd

Fix missing return statement for test_it_returns_empty_list_when_model_contains_supported_inplace_ops in #89299 (#90797)

kevalmorabia97 committed 2 years ago

e90169d1

[FSDP][2/N] Refactor state dict hook registration (#90777)

awgu committed 2 years ago

510339c0

Small fixes for better channels last performance (#89616)

Natalia Gimelshein committed 2 years ago

ed050e7a

Older