strisunshinewentingwang/pytorch

Pull Requests Commits

CUDA fast path for `_chunk_cat()` (#120678)

BoyuanFeng committed 189 days ago

aaff8d27

allow fp16 in quant/dequant decompositions (#121738)

manuelcandales committed 189 days ago

c53e3f57

[DDP][PT2D][2D] Enable DDP + TP and add test for compiled DDP + TP (#120479)

fegin committed 189 days ago

c7193f40

[Export, AOTInductor] Populate ShapeEnv's var_to_val during deserialization (#121759)

SherlockNoMad committed 189 days ago

dd568f42

Revert "Init CUDA instead of faking memory stats (#121698)"

pytorchmergebot committed 189 days ago

a2a4693c

Revert "[compiled autograd] free stack objects before calling compiled graph (#121707)"

pytorchmergebot committed 189 days ago

45a835ce

[compiled autograd] support custom ops backed by c++ autograd::Function (#120681)

xmfan committed 189 days ago

8b1b61bc

Add support for tt.scan to triton kernel mutation analysis (#121828)

oulgen committed 189 days ago

58ff55aa

[DDP][PT2D] Allreduce fusion fx pass using concat and all_reduce_coalesced (#113209)

fegin committed 189 days ago

8e6d572b

Support `call_method` in DDPOptimizer (#121771)

BoyuanFeng committed 189 days ago

0c1ac448

[dynamo] Convert invalid args into graph breaks (#121784)

jansel committed 189 days ago

0df39480

[compiled autograd] free stack objects before calling compiled graph (#121707)

xmfan committed 189 days ago

5b900745

Init CUDA instead of faking memory stats (#121698)

xmfan committed 189 days ago

2460f0b1

Support setUpClass & tearDownClass with instantiate_device_type_tests() (#121686)

masnesral committed 189 days ago

cd949d13

Count the number of entries directly in avg_pool2d lowering (#121429)

isuruf committed 189 days ago

ffabb25c

Add lowering for avg_pool{1, 3}d (#116085)

isuruf committed 189 days ago

a19a05fd

Use pytorch bot's labeler (#121762)

clee2000 committed 189 days ago

79fac48b

Allow custom attributes for torch function subclasses (#121693)

mlazos committed 189 days ago

05df03ec

Make translation validation more user friendly (#120880)

ezyang committed 189 days ago

92a2b214

Upgrade to tlparse 0.3.7 (#121772)

ezyang committed 189 days ago

b1d59989

[MPS] Fix naive matmul for BFloat16 (#121731)

malfet committed 189 days ago

5498804e

[dynamo] Refactor TorchInGraphFunctionVariable for compile time (#121616)

jansel committed 189 days ago

559ca13b

Revert "Include torch warn in each error in cudnn/Conv_v8.cpp (#120719)"

pytorchmergebot committed 189 days ago

51cf57c6

[constraints] Fix scalar type for constraint_range to Long (#121752)

IvanKobzarev committed 189 days ago

a157a0d0

make _process_dynamic_shapes an implementation detail (#121713)

avikchaudhuri committed 189 days ago

7fe0cc53

Add quantized conv transpose2d op (#120151)

shubhraprakash1 committed 189 days ago

5088e495

Back out "[DeviceMesh] Add support for nD slicing (#119752)" (#121763)

wz337 committed 189 days ago

e99fa004

add std::ostream& operator<< for BFloat16 in BFloat16.h (#121302)

LamForest committed 189 days ago

be33d31a

[nit][DCP][DSD] Remove variables not being used in test_state_dict.py #121204 (#121773)

wz337 committed 189 days ago

5986552e

`_foreach_copy` with different src/dst dtypes (#121717)

crcrpar committed 189 days ago

da2a9a05

Newer Older