strisunshinewentingwang/pytorch

Pull Requests Commits

[MPS] Fix gradient issues with NLL and Smooth_L1 loss ops (#94226)

razarmehr committed 2 years ago

368e364c

Some more fixes (#94049)

cyyever committed 2 years ago

bf9be50b

Revert "enable bf16 emb (#94163)"

pytorchmergebot committed 2 years ago

53e4fe07

Look up `group["capturable"]`, not `defaults["capturable"]` in Adam(W) (#94149)

crcrpar committed 2 years ago

6ba041fc

Cleanup all leftover processes in MacOS pet runner (#94127)

huydhn committed 2 years ago

0dfc3e13

[inductor] Avoid re-computing mean in lowering for aten.var_mean (#94139)

peterbell10 committed 2 years ago

a595d06c

[inductor] Count bytes can't read from buffers that are never written (#94142)

peterbell10 committed 2 years ago

719f78d3

Extend torch-trition conda to 3.11 (#93117)

malfet committed 2 years ago

43f6ed4a

use C10_API in libc10.so (#94171)

cyyever committed 2 years ago

3c6bc58f

Re-enable compilation tests (#92333)

malfet committed 2 years ago

a07d1291

Fix bug in generic_list_compare (#94156)

jansel committed 2 years ago

180adf8c

Point to scatter_reduce for reduce argument in scatter_ docs (#94081)

mikaylagawarecki committed 2 years ago

fdebc062

Make linter quick-checks setup steps retryable (#94199)

ZainRizvi committed 2 years ago

05397b12

Make segment_reduce properly private. (#93166)

albanD committed 2 years ago

496c0a20

Make sure to properly pull the right submodule in BC test (#94182)

albanD committed 2 years ago

9b3277c0

Revert "Support neg calls to dyn shapes (#94068)"

pytorchmergebot committed 2 years ago

0444b8f5

[Inductor] Performance smoke test - hf bert performance increased (#94088)

weiwangmeta committed 2 years ago

9b2e7d3b

Don't compare ids of temporary python objects (#94097)

albanD committed 2 years ago

d2b82feb

Fix serialization (#94096)

albanD committed 2 years ago

25a6e0fd

Skip sebotnet33ts_256 on CI (#94067)

desertfire committed 2 years ago

db011e11

[DCP] Fix test_file_system_checkpoint.py and test_file_system_checkpoint_cpu.py (#94069)

wz337 committed 2 years ago

16387bee

[decomp] Decompose std/std_mean into aten.var/var_mean (#94072)

peterbell10 committed 2 years ago

819990f5

Optimize ConvTransposed2D with mkldnn float32 and bfloat16 on CPU (#92530)

mingfeima committed 2 years ago

26cba842

enable bf16 emb (#94163)

zhuhaozhe committed 2 years ago

f3bf46e8

fix inductor clamp decomp to correctly type promote and avoid wrappin… (#94157)

Natalia Gimelshein committed 2 years ago

ea4cda52

Support neg calls to dyn shapes (#94068)

voznesenskym committed 2 years ago

9350bcf6

Add missing move to torch_dispatch_mode.h (#94154)

Skylion007 committed 2 years ago

7b6e9488

[MPS] Fix `cumsum` for negative indexes (#94119)

malfet committed 2 years ago

10a1efb4

Small refactor of shape guards to allow for 1:1 code_parts (#93894)

voznesenskym committed 2 years ago

60a3b742

[MPS] Fix `index_select` for empty input (#94117)

malfet committed 2 years ago

8a88852d

Older