strisunshinewentingwang/pytorch

Pull Requests Commits

[FSDP] Optimizer states may be on CPU, copy them to GPU before gathering (#84708)

fegin committed 2 years ago

28c830ac

Missed one CHECK_NOTNULL in #82032's find-replace (#84720)

cchan committed 2 years ago

0fd8f6b9

Add user facing documentation for CSAN (#84689)

sypneiwski committed 2 years ago

d12f3524

remove c10_defs.bzl and embed its logic directly where it is used (#84595)

Michael Andreas Dagitses committed 2 years ago

a8198a09

[quant][docs] Additonal fixes for quantize_fx docs (#84587)

jerryzh168 committed 2 years ago

214a6500

Set up aten/src/ATen/functorch directory; move some files there (#84648)

zou3519 committed 2 years ago

0a89bdf9

Add CSAN support for CPU synchronizations (#84428)

sypneiwski committed 2 years ago

8e57ce63

[functorch] Simplify BatchedTensorImpl (#84642)

zou3519 committed 2 years ago

7702ca49

[Mobile] Use -ffunction-sections and -fdata-sections for Mobile builds (#84704)

dhruvbird committed 2 years ago

0d46bfac

[Mobile] Update build_mobile.sh to allow lite interpreter and tracing based builds (#84647)

dhruvbird committed 2 years ago

747f27a9

[DataPipe] Fix mishandling of exception message when error is not iterable (#84676)

NivekT committed 2 years ago

27e5299e

[functorch] delete functorch/csrc/Constants.h (#84639)

zou3519 committed 2 years ago

df3377fb

ROCm support for test_lazy_init (#84333)

jataylo committed 2 years ago

09bcc006

Add synchronize hooks (#84427)

sypneiwski committed 2 years ago

67d6f716

Redo how custom/python_custom methods on TensorImpl work (#84641)

ezyang committed 2 years ago

591b75bf

Add config to PrimTorch's nvFuser executor (#84482)

IvanYashchuk committed 2 years ago

d802fcfc

test mkldnn conv2d channels last when weight is nchw format (#77348)

mingfeima committed 2 years ago

6f72c13f

[FSDP] Ensure that all ranks use the same order to iterate through optimizer states (#84654)

fegin committed 2 years ago

1840f24d

Moving CommTensor from tests to private _spmd folder (#84719)

mrshenli committed 2 years ago

22119495

[torch.fx.wrap] Use callable / function.__name__ instead of function.__code__.co_name (#84373)

kunalb committed 2 years ago

b00a4b7c

[vision hash update] update the pinned vision hash (#84731)

pytorchmergebot committed 2 years ago

1c8cb087

[torchdynamo hash update] update the pinned torchdynamo hash (#84730)

pytorchmergebot committed 2 years ago

1c8f02d4

Return x.alias() when transpose is an nop (#84674)

SherlockNoMad committed 2 years ago

e6ee8e61

[ONNX] Fix node attributes when namespace is aten (#84211)

justinchuby committed 2 years ago

dbdc1cd5

[ONNX] Rename constants for clarity (#84645)

justinchuby committed 2 years ago

2fa8142c

[quant] remove mkldnn headers in OnednnUtils.h (#84195)

xndcn committed 2 years ago

bc3683de

Disable Transformer/MHA fast path when autocast is enabled (#84722)

erichan1 committed 2 years ago

7a5d5a00

Fix conda cmake setup for macos x86-64 (#84682)

huydhn committed 2 years ago

f23a1cf8

Add __all__ for a few distributed modules plus a little typing (#84119)

Rodrigo Kumpera committed 2 years ago

6f216805

[functorch] excise older custom_vjp prototype (#84638)

zou3519 committed 2 years ago

9b8e0a38

Newer Older