strisunshinewentingwang/pytorch

Pull Requests Commits

Plumb various exporter strategies through new API

abock committed 1 year ago

f8ed7cc6

Introduce torch.onnx.dynamo_export API

Thiago Crepaldi committed 1 year ago

7d919fc2

Add parentheses to `FloorDiv`. (#98290)

ysiraichi committed 1 year ago

42cbf712

[FSDP] Use correct handle training state when prefetching (#98249)

awgu committed 1 year ago

0b31f87c

extract out a caffe2 macros library (#98156)

mikey dagitses committed 1 year ago

950431c3

[FSDP] Allow non-uniform `requires_grad` for `use_orig_params=True` (#98221)

awgu committed 1 year ago

f6272ce7

generate caffe2/core/macros.h in shared build structure (#98131)

mikey dagitses committed 1 year ago

301f00f3

Align settings for new device key. (#98224)

egienvalue committed 1 year ago

d47a4bf5

Disable inductor/test_minifier on ASAN (#98263)

huydhn committed 1 year ago

86505c69

fix the use of incomplete vector<T> for C++20 compatibilities (#93978)

ppwwyyxx committed 1 year ago

e7874eea

[Dynamo] Support skip fbcode modules (#98192)

yanboliang committed 1 year ago

a9c7e882

[inductor] be able to enable max-autotune and cudagraphs independently (#98255)

shunting314 committed 1 year ago

d16a9b76

Revert "Extend TensorImpl with BackendMeta (#97429)"

pytorchmergebot committed 1 year ago

7eaaefaf

[torch/fx] add torch/utils/_stats.py to stack frame skiplist (#98117)

suo committed 1 year ago

8f2f1a0b

add support for SymNodeVariable in getitem_const (#97756)

blzheng committed 1 year ago

1fae179e

[quant][pt2e][refactor] Remove `backend_config` from `_maybe_insert_input_observers_for_node` (#98094)

jerryzh168 committed 1 year ago

b1090830

Extend TensorImpl with BackendMeta (#97429)

ppiskorski committed 1 year ago

bc38b278

[vision hash update] update the pinned vision hash (#98261)

pytorchmergebot committed 1 year ago

c5963b77

introduce c10::DataPtr::mutable_get() and use it in c10 (#98217)

mikey dagitses committed 1 year ago

4431509a

Revert "Add all_reduce_coalesced functional collective (#97157)"

pytorchmergebot committed 1 year ago

fa08e546

[inductor] [cpp] fix bitwise codegen (#98056)

Valentine233 committed 1 year ago

177994eb

Inductor cpp wrapper: support LinearUnary (#97655)

chunyuan-w committed 1 year ago

0f151ad2

Create script to upload test aggregation data (#97954)

PaliC committed 1 year ago

0e2bde30

[dynamo benchmarks] Fix inference benchmark runs (#98248)

williamwen42 committed 1 year ago

4cf3e7c2

Added ModuleInfos for {*}Norm modules (#97919)

mikaylagawarecki committed 1 year ago

96ad739d

Add all_reduce_coalesced functional collective (#97157)

Rodrigo Kumpera committed 1 year ago

a3fc3531

Skip gat, gcn and sage for TorchBench CUDA test (#98244)

desertfire committed 1 year ago

69ff39d2

[PyTorch] Don't do extra numel() check in TensorImpl::data() (#98090)

swolchok committed 1 year ago

f386312e

Switch reduce_scatter and all_gather in DeviceMesh to use functional collectives (#96226)

Rodrigo Kumpera committed 1 year ago

9ad66dd5

run buildifier on unified build files (#98141)

mikey dagitses committed 1 year ago

2ac90869

Older