strisunshinewentingwang/pytorch

Pull Requests Commits

[MPS] Fix the argument error for tensor_split() test (#94234)

razarmehr committed 1 year ago

b654d149

[MPS] Remove the unused code for view lists in OperationUtils.h (#94265)

razarmehr committed 1 year ago

a3ca66c6

[MPS] Don't reset the Graph state (#94283)

kulinseth committed 1 year ago

a0a37280

[MPS] Fix the crash in View ops when slicing wrong lengths (#94259)

razarmehr committed 1 year ago

36062dd2

General in-place binary op support in dynamo (#94203)

jbschlosser committed 1 year ago

bf4fe5dd

Dynamo: Fix to unpack ConstantVariable in call_range() (#94202)

jbschlosser committed 1 year ago

f954498e

Fix thread-allocation in `_vec_log_softmax_lastdim` (#85398)

sanchitintel committed 1 year ago

c4544bc1

update test fixture (#89796)

eellison committed 1 year ago

a2ac25f6

sparse compressed tensor validation without syncs for low-(batch)dim tensors. (#94048)

nikitaved committed 1 year ago

513b5da3

[BE] Add empty tensor check to _compute_linear_combination (#94245)

malfet committed 1 year ago

42b6bcdb

[Inductor] Fix CPU vectorized implementation of mask calculation that breaks torch.where (#93922)

jgong5 committed 1 year ago

a28a0629

[inductor] bug fix: use `create_symbolic_sizes_strides_storage_offset` (#94031)

nkaretnikov committed 1 year ago

0e94fbc0

[Dynamo] Support torch.Tensor.fn as TorchVariable, not UserDefinedObjectVariable, preventing graph break (#93243)

min-jean-cho committed 1 year ago

900e09c8

Refactor sharding data pipe into a seperate file (#94095)

wenleix committed 1 year ago

d6dec1a5

[quant][fx][pt2e] Refactor prepare so it's aligned better with the new API plan in pt2e (#94011)

jerryzh168 committed 1 year ago

59c1b502

[Docs] Add pointer to FlashAttention paper (#94253)

tridao committed 1 year ago

ffb3561c

Clean up mentions of removed torch/csrc/generic/*.cpp (#94107)

kit1980 committed 1 year ago

f92348e1

[MPS] Unregister put_() op due to lack of implementation (#94231)

razarmehr committed 1 year ago

bc8a3783

[FSDP][optim_state_dict] Let optim_state_dict ignore the non-FSDP managed parameters that do not reside on the rank (#94129)

fegin committed 1 year ago

bc6d54f6

[FSDP][state_dict] Fix incorrect valid_data_size for local_state_dict when some ranks have zero data. (#94109)

fegin committed 1 year ago

f04106f1

FakeTensor should constant propagate through ops that allow numbers as scalars (#94145)

yanboliang committed 1 year ago

605b6618

[mobile] List all missing ops at once (#94205)

larryliu0820 committed 1 year ago

579ae64d

Use official NVML Python bindings (#93925)

XuehaiPan committed 1 year ago

4b0e2e2c

Revert "Add fabi-version=11 to ensure compatibility between gcc7 and gcc9 binaries for _GLIBCXX_USE_CXX11_ABI=1 (#93835)"

pytorchmergebot committed 1 year ago

10633948

[vision hash update] update the pinned vision hash (#94241)

pytorchmergebot committed 1 year ago

f1c435d7

Add fabi-version=11 to ensure compatibility between gcc7 and gcc9 binaries for _GLIBCXX_USE_CXX11_ABI=1 (#93835)

zhuhong61 committed 1 year ago

b562be79

[MPS] Add scalar params to the softplus key. (#94256)

kulinseth committed 1 year ago

ca741053

[MPS] Handle empty input in layer norm (#94212)

DenisVieriu97 committed 1 year ago

9358726a

[MPS] Return input in addcmul/div if value is zero (#94214)

razarmehr committed 1 year ago

d493bc8a

[MPS] Fix the crash in nan_to_num() with Float16 data type (#94220)

razarmehr committed 1 year ago

fa2b99f4

Newer Older