strisunshinewentingwang/pytorch

Pull Requests Commits

Add invariant that all symbolic shapes must be bound in graph (#99089)

ezyang committed 1 year ago

039faf0d

[SPMD][BE] Generalize factory ops support in SPMD expansion (#99233)

mrshenli committed 1 year ago

c69d5488

[SPMD][BE] Remove deprecated aten.sym_numel branch (#99232)

mrshenli committed 1 year ago

6bb20822

[SPMD][BE] Consolidate DSymInt Branches (#99231)

mrshenli committed 1 year ago

39be9949

[SPMD] Refactor DSize to DSymInt to enable sym_numel (#99206)

mrshenli committed 1 year ago

544cd8e1

[SPMD] Enable aten.full.default with SymInt on sharded dims (#99190)

mrshenli committed 1 year ago

bafb9840

SymIntify randint and randperm (#98968)

ezyang committed 1 year ago

d350646f

Support large negative SymInt (#99157)

ezyang committed 1 year ago

756a86d5

[vmap] Add vmap support for nn.functional.huber_loss (#99235)

qqaatw committed 1 year ago

5c062e8b

fix breakage from #99027 (#99245)

mikey dagitses committed 1 year ago

c9403f12

Update NT error message (#99166)

cpuhrsch committed 1 year ago

3cde50e3

[dynamo] Support DELETE_ATTR (#98698)

jansel committed 1 year ago

47c685de

[Dynamo] Fix benchmark --verbose error (#99224)

yanboliang committed 1 year ago

15fe5a07

[pt2] add `SymInt` support for fft ops (#99115)

nkaretnikov committed 1 year ago

21681f36

[pt2] add `SymInt` support for `roll` (#99114)

nkaretnikov committed 1 year ago

f89b7c2b

Apply dynamic shapes policy correctly to _base tensor (#99211)

voznesenskym committed 1 year ago

d5f7ec8a

[BE] Update flake8-comprehensions and adapt to rule C418 (#99178)

Skylion007 committed 1 year ago

85f38b8a

make ATen/native/cuda/NLLLoss2d.cu data_ptr-correct (#99179)

mikey dagitses committed 1 year ago

506bd057

make ATen/native/cuda/AdaptiveMaxPooling3d.cu data_ptr-correct (#99185)

mikey dagitses committed 1 year ago

e9201ab6

[CI] Remove inductor skip list for timm_models (#98840)

desertfire committed 1 year ago

34f681c1

[CI] Use expected accuracy csv files to check benchmark test status (#98839)

desertfire committed 1 year ago

a595a506

nn.Linear: dispatch to bsr_dense_mm for half and bfloat16 (#94825)

nikitaved committed 1 year ago

1adb6fa9

make ATen/native/cuda/MaxUnpooling.cu data_ptr-correct (#99189)

mikey dagitses committed 1 year ago

b69f0480

make ATen/native/cuda/UpSampleNearest2d.cu data_ptr-correct (#99186)

mikey dagitses committed 1 year ago

60f914e7

[Dynamo] No graph break for explicit calling Conv{1/2/3}d.forward & ConvTranspose{1/2/3}d.forward (#99015)

yanboliang committed 1 year ago

05809c7d

Enable FSDP ``use_orig_params=True`` mixed precision training when some ranks have no (non-zero sized) parameter shards (#99175)

speediedan committed 1 year ago

157c8690

[dynamo] Support functools.wraps (#98699)

jansel committed 1 year ago

e9be0b0f

[vision hash update] update the pinned vision hash (#99212)

pytorchmergebot committed 1 year ago

b9426ded

Patch failing slow-test logic for inductor-dynamic (#99182)

wconstab committed 1 year ago

3c4622c0

Graph-break on allowed modules if they have hooks (#97184)

wconstab committed 1 year ago

6eab5e88

Newer Older