strisunshinewentingwang/pytorch

Pull Requests Commits

[c10d] Start deprecating *_multigpu APIs (#85961)

kwen2501 committed 1 year ago

05d11281

[c10d] Start deprecating *_coalesced APIs (#85959)

kwen2501 committed 1 year ago

463283e0

Fix the error with constant_pad_nd for 4D+ padding (#85991)

razarmehr committed 1 year ago

bf667c63

[FSDP] Ignore buffers that are non-persistent. (#85740)

fegin committed 1 year ago

be29ca97

Revert "[maskedtensor] use masked_softmax for forward/backward instead of regular softmax (#85845)"

pytorchmergebot committed 1 year ago

db4c6fe5

Refactored recomputable ops a bit and added a bunch more ops (#85993)

Chillee committed 1 year ago

9bf9db57

Removed debug output that doesn't work with faketensors (#85992)

Chillee committed 1 year ago

e09a84a1

[Quant] Make x86 backend default when querying qconfig (#85461)

Xia-Weiwen committed 1 year ago

4b86a935

nvprim op support runtime checks on dtype compatibility on prims.convert_element_type (#85566)

jjsjann123 committed 1 year ago

fd553c46

[BE] Get rid of `std::result_of` in `c10` (#85977)

malfet committed 1 year ago

01292cc9

Fix fake tensor kernel nesting (#85920)

eellison committed 1 year ago

c2d9ea7f

Lazily load decompositions for jvp (#85989)

soulitzer committed 1 year ago

28061d50

Fix the dimension of padding to match the input's dimension (#85990)

razarmehr committed 1 year ago

334686bd

[Quant] Enable XNNPACK ops in QNNPACK BackendConfig (#85863)

andrewor14 committed 1 year ago

24fc680e

added fix for WorkUCC (#84368)

Fuzzkatt committed 1 year ago

d9421f81

[MPS] Enable caching for random ops with Philox engine (#85833)

razarmehr committed 1 year ago

a4cc6399

[quant] Fix per channel weight observer (#85883)

digantdesai committed 1 year ago

071f8750

[test_nn] split embedding tests from test_nn (#85892)

kshitij12345 committed 1 year ago

6a5550fc

Make FunctionalTensorWrapper correctly handle symbolic shapes (#85975)

ezyang committed 1 year ago

2037b7cb

Consistent compute numel/contiguous strategy with SymInts (#85858)

ezyang committed 1 year ago

3b6588ab

Enable convolution_backward with bias and symints (#85970)

ezyang committed 1 year ago

84a06d71

[maskedtensor] use masked_softmax for forward/backward instead of regular softmax (#85845)

george-qi committed 1 year ago

a4d10342

[BE] Generate names of known device from array (#85982)

malfet committed 1 year ago

1c970846

Revert "[CUBLAS][CUDA GRAPHS] (re-re-open of #83461) Explicitly set the workspace for cuBLAS handles (#85447)"

pytorchmergebot committed 1 year ago

71eb0440

[ci] two procs for parallelization (#85985)

clee2000 committed 1 year ago

401a3588

[functorch] test no warning on `import functorch` (#85980)

zou3519 committed 1 year ago

e73e3e35

[functorch] fix whirlwind tour ipynb (#85974)

zou3519 committed 1 year ago

eed5f046

Remove `optim_mt` from `test/test_optim.py` (#83549)

crcrpar committed 1 year ago

4c04fa95

LU solve/unpack fix to prevent bad memory usage on CPU (#85922)

albanD committed 1 year ago

94da90e4

Disallow saved tensor hooks in functorch transforms (#85972)

zou3519 committed 1 year ago

7238ca4c

Newer Older