pytorch/xla

Pull Requests Commits

Wire in dense gather.

dlibenzi committed 6 years ago

4d38f69a

Bring the latest TF code with multi TCP stream support and new torch gather.

dlibenzi committed 6 years ago

d2a827c1

Implement client side cluster resolving (#910)

jysohn23 committed 6 years ago

Verified 0d0f3f8d

Make nll_loss honor ignore_index

asuhan committed 6 years ago

68704a54

Merge pull request #906 from ailzhang/fix_pt_23790

ailzhang committed 6 years ago

Verified 5c2d42b5

reorder alphabetically

Ailing Zhang committed 6 years ago

d7cccead

Ailing Zhang committed 6 years ago

5fd0e91d

fix breakage #23790

Ailing Zhang committed 6 years ago

45e3a4a0

Implement cluster validation (#875)

jysohn23 committed 6 years ago

Verified 16ea1f1a

Install `coverage` package in docker (#874)

jysohn23 committed 6 years ago

Verified 418fd683

Merge pull request #833 from pytorch/asuhan/bool_comparison_type

ailzhang committed 6 years ago

Verified 5d4fcad6

Make tensor comparison return type match upstream PyTorch

asuhan committed 6 years ago

74c98bb9

Define classes to be used for distributed training (#864)

jysohn23 committed 6 years ago

Verified 345522b6

dont crash with one core when checking replication (#854)

zcain117 committed 6 years ago

Verified 229d0e2d

Merge pull request #858 from pytorch/validation_step

taylanbil committed 6 years ago

Verified 3f5b62a1

Merge pull request #857 from taylanbil/dpapi

taylanbil committed 6 years ago

Verified 6a23e110

updated api per offline discussion

taylanbil committed 6 years ago

e1e64a10

Add tensor's barrier at the DataParallel iterator level, to make sure that even during validation (which is missing the optimizer step call), tensor's are synced.

dlibenzi committed 6 years ago

18c13e8f

added a models_dict to DataParallel

taylanbil committed 6 years ago

9a56a152

Add environment configuration for waiting on the tensors sync operation.

dlibenzi committed 6 years ago

0fbb2518

Add performance comment on the line of the one suggested by Zach.

dlibenzi committed 6 years ago

deb652ad

Merge pull request #852 from pytorch/repli_full_devices

taylanbil committed 6 years ago

Verified afa160a2

Replication is only for the full device set at this time.

dlibenzi committed 6 years ago

aa9153c4

Added item() call advise.

dlibenzi committed 6 years ago

b76b0242

Remove the file based XRT config to uniformly use the environment based configuration.

dlibenzi committed 6 years ago

4d5f9fe3

Eval loader transforms (#850)

zcain117 committed 6 years ago

Verified 5034ebf2

Refresh gradient clip patch.

dlibenzi committed 6 years ago

41dc90e0

Handle different GDB stack trace formats.

dlibenzi committed 6 years ago

c7ac1173

Define explicit entrypoint for container (#842)

jysohn23 committed 6 years ago

Verified 4d2f1e70

Allow metrics to be saved into a specified file.

dlibenzi committed 6 years ago

5166457a

Older