pytorch/xla

Pull Requests Commits

Allow iterator regeneration.

dlibenzi committed 6 years ago

6786b082

Add logging to our passes

asuhan committed 6 years ago

e667d2f6

Change the readme to point to newest PyTorch

asuhan committed 6 years ago

9101e295

Changed the computation client interface to be compile handle based to get around the continuous marshaling of computation protos.

dlibenzi committed 6 years ago

1de9c055

Allow an XRT cached node to have multiple outputs/operations.

dlibenzi committed 6 years ago

e842b9fa

Added ability to run train session with fake data.

dlibenzi committed 6 years ago

07de6585

Simplify APIs and remove redundant logging

asuhan committed 6 years ago

0288df9a

Address more review feedback

asuhan committed 6 years ago

e0f13451

Fix the construction of argument list for backward

asuhan committed 6 years ago

e9248d3c

Even more cleanups from review feedback

asuhan committed 6 years ago

3be22fed

Additional XlaModule fixes per review feedback

asuhan committed 6 years ago

5f564015

Store gradient descriptor in XlaModule

asuhan committed 6 years ago

5cbc0aa4

Additional cleanups from PR feedback

asuhan committed 6 years ago

f5b5faa9

Adjust signature of RemoveUnusedForwardOutputs

asuhan committed 6 years ago

80a8da60

Address some of the logging PR feedback

asuhan committed 6 years ago

6b217690

Add and use XlaHelpers::SizesOfXlaOp helper

asuhan committed 6 years ago

02832684

Remove nearly all dependencies on PyTorch trace shapes

asuhan committed 6 years ago

842ca03b

Add pass to replace inplace operators with non-inplace counterparts

asuhan committed 6 years ago

e849a657

Add pass to remove in-place operations on output parameters

asuhan committed 6 years ago

4b1edad5

Keep track of evaluated aten::size

asuhan committed 6 years ago

7ded34a6

Initial changes for latest PyTorch

asuhan committed 6 years ago

abd38b35

Merge pull request #62 from pytorch/test_util_restructure1

jysohn23 committed 6 years ago

Verified d4fbf359

Do not import xla_model.py from test_util.py, as this ends up picking up the xla_model.py from the installed module, instead of the local one.

dlibenzi committed 6 years ago

b709d55e

Add tensorboardX support for all our models

jysohn23 committed 6 years ago

Verified 161686d9

Addressed review comments and ran google-pyformat

jysohn23 committed 6 years ago

cd35dc52

Handle cases for metrics_debug

jysohn23 committed 6 years ago

ea068be1

Remove explicit writer from cifar and imagenet models

jysohn23 committed 6 years ago

b3e9cfcf

Fix processed examples calculated

jysohn23 committed 6 years ago

7ec1597c

Create new classes for writing and logging metrics

jysohn23 committed 6 years ago

1d410082

Merge branch 'jysohn/tbx' of https://github.com/pytorch/xla into jysohn/tbx

jysohn23 committed 6 years ago

016b6aa2

Older