pytorch/xla

Pull Requests Commits

Allow the loop functions to save device specific data (ie, optimizers, in case of them carrying LR schedules) across training loop.

dlibenzi committed 6 years ago

dd52ee5f

Revert "Revert "Fix brekages from recent PT changes.""

dlibenzi committed 6 years ago

dd9d330e

taylanbil committed 6 years ago

dc756789

build wheels script bug fix

taylanbil committed 6 years ago

87c965c2

Enable running model on PyTorch CPU.

dlibenzi committed 6 years ago

5e1b0c9e

Fix the simple scalar detection logic. (#795)

dlibenzi committed 6 years ago

7b1e33d0

Revert "Fix brekages from recent PT changes."

dlibenzi committed 6 years ago

c795d861

Make dropout pass-through in !train mode.

dlibenzi committed 6 years ago

bc8f03c0

add a testcase for printing tensors on xla

Ailing Zhang committed 6 years ago

806920e3

Merge pull request #787 from pytorch/fixreplimode

taylanbil committed 6 years ago

Verified e6a4b673

Fix data-parallel single single device run.

dlibenzi committed 6 years ago

58d730fe

Fix brekages from recent PT changes.

Ailing Zhang committed 6 years ago

5d9c7d73

_get_model_device should not return None

taylanbil committed 6 years ago

8ac4d16f

Clean up data parallel and remove unused test API.

dlibenzi committed 6 years ago

2e35c6b8

Remove unused ATen/Type.h header

asuhan committed 6 years ago

12527a8c

Simplify gradient collection, since we do not need tensors anymore.

dlibenzi committed 6 years ago

e9876950

Minor fixes in the README (#778)

vishwakftw committed 6 years ago

9a6ec060

Add import of os in CIFAR training test.

dlibenzi committed 6 years ago

00c4c7b2

Add release_version variable for stable builds

jysohn23 committed 6 years ago

52a5e102

Plug a default device setting if a user asks for a single device.

dlibenzi committed 6 years ago

d03f827b

Use proper where mapping.

dlibenzi committed 6 years ago

29306d8f

Update update_torch_wheel script (#772)

jysohn23 committed 6 years ago

Verified 97b42695

Fix the Compile A Palooza issue that arise after the PT Variable code change.

dlibenzi committed 6 years ago

c0a7caea

Override ATEN to() operations.

dlibenzi committed 6 years ago

0768145b

Fix aten_xla_type.h formatting

asuhan committed 6 years ago

ca378226

Update README.md

jysohn23 committed 6 years ago

7c3091cc

Clarify the required bazel version

apaszke committed 6 years ago

dbb11d00

Merge pull request #766 from ailzhang/fix_pt20994

ailzhang committed 6 years ago

Verified 49a20afe

fix breakage PT#20994

Ailing Zhang committed 6 years ago

12c3a7b3

Merge pull request #764 from pytorch/kaboom_rebased

ailzhang committed 6 years ago

Verified f4bd9cc2

Older