pytorch/xla

Pull Requests Commits

Fix batch normalization to be able to deal with train() and eval() modes.

dlibenzi committed 6 years ago

e5659f86

Fix C++ tests warnings.

dlibenzi committed 6 years ago

c1ae6801

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

b1f944ac

Merge pull request #804 from ailzhang/refactor_cpp_test

ailzhang committed 6 years ago

Verified ef4ff9ff

Ailing Zhang committed 6 years ago

52ba56f7

Merge branch 'master' of https://github.com/pytorch/xla into refactor_cpp_test

Ailing Zhang committed 6 years ago

49abdbc0

use oss config for clang-format

Ailing Zhang committed 6 years ago

c0f348d4

search & replace with a wrapper to .to(..., /*non_blocking=*/, /*copy=*/)

Ailing Zhang committed 6 years ago

32e7863b

use AtenDefaultDevice

Ailing Zhang committed 6 years ago

87a9f6aa

ForEachTorchXLADevice->ForEachDevice

Ailing Zhang committed 6 years ago

5b53bfd3

refactor cpp tests

Ailing Zhang committed 6 years ago

44a7c3c6

Enable at::expand() to understand -1 expand dimensions.

dlibenzi committed 6 years ago

cab477c2

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

Older