pytorch/xla

Pull Requests Commits

simple version of high-accuracy resnet50 learning schedule

zcain117 committed 6 years ago

c1224ca0

Merge branch 'master' of https://github.com/pytorch/xla into add-imagenet-learning-schedule

zcain117 committed 6 years ago

8e0ef45e

simple learning rate scheduler for resnet50'

zcain117 committed 6 years ago

32dea78f

punt back to cpu for xla::dot related ops with int input (#999)

ailzhang committed 6 years ago

a98273a6

add comment about half (#998)

ailzhang committed 6 years ago

Verified 3cb6e972

better error message for arange (#1002)

ailzhang committed 6 years ago

9f6a18ee

Make sign and abs work with unsigned integers

asuhan committed 6 years ago

45507ac6

Shard datasets for distributed training. (#997)

jysohn23 committed 6 years ago

Verified 6341b5c0

add global rate and wall time to train print statement (#988)

zcain117 committed 6 years ago

Verified 582f612f

address comment

Ailing Zhang committed 6 years ago

195fd9a3

fix test; address comment

Ailing Zhang committed 6 years ago

8a73b1ad

check in call site

Ailing Zhang committed 6 years ago

a1629865

allow mean return nan for 0dim

Ailing Zhang committed 6 years ago

a8c8fff2

add test for 0dim reduction ops

Ailing Zhang committed 6 years ago

c6655607

Fix descending sort when minimum values are present

asuhan committed 6 years ago

e0eff17a

fix pow for integral types (#994)

ailzhang committed 6 years ago

Verified f7fc05ab

zcain117 committed 6 years ago

1640784e

move add_Scalar to test_utils

zcain117 committed 6 years ago

374ac858

only write if logdir is not None

zcain117 committed 6 years ago

f6cd2ef4

zcain117 committed 6 years ago

a1cc68b6

default log_dir to FLAGS.logdir

zcain117 committed 6 years ago

24a7b1ae

add percent sign

zcain117 committed 6 years ago

30889dd5

print mean accuracy

zcain117 committed 6 years ago

7f9ed874

add tensorboard support

zcain117 committed 6 years ago

660e9331

Add comment regarding tensor stride

asuhan committed 6 years ago

fc7e052d

fix split with 0dim (#991)

ailzhang committed 6 years ago

Verified 54cd2534

Fix the XLA tensors updates back from CPU tensors, in case of writeable tensors.

dlibenzi committed 6 years ago

d0e5be7b

Cleaup Kokoro CI infra bits. (#987)

jysohn23 committed 6 years ago

c469575d

Fix elementwise arithmetic with zero sized dimensions

asuhan committed 6 years ago

a1b571cb

raise error for reduction ops when numel==0 (#981)

ailzhang committed 6 years ago

Verified 61fe256f

Older