365 tensorflow upgrade nn hack

danieltudosiu commented 5 years ago

Status

READY

Description

NiftyNet was tested if it works with Tensorflow 1.13.1/1.14.0rc0 and CUDA 10.0.130 with cudnn 7.5.

Types of changes

[x] New feature (non-breaking change which adds functionality)
[x] My code follows the code style of this project.
[x] All new and existing tests passed.

Todos

[x] Tests
[x] Documentation

Impacted Areas in Application

List general components of the application that this PR will affect: niftynet	-layer
-crf.py

-contrib
|   |
|   -segmentation_selective_sampler
|       |
|       -test_sampler_selective.py
-io
    |
    -misc_io.py

tests | -most tests (too lazy to put 25+ files in the nice format I used)

danieltudosiu commented 5 years ago

@ericspod what are we doing with this pull request?

danieltudosiu commented 5 years ago

So what should we do based on the Tensorflow Issue Unittest and test_session interaction #29439 response? Change the tests naming scheme so we do not use the non-OOP fix?

ericspod commented 5 years ago

It wasn't our names it was theirs that was the issue. I say we keep our delattr solution for now.

Eric

On Thu, 13 Jun 2019, 14:59 Petru-Daniel Tudosiu <notifications@github.com wrote:

So what should we do based on the Tensorflow Issue Unittest and test_session interaction #29439 https://github.com/tensorflow/tensorflow/issues/29439 response? Change the tests naming scheme so we do not use the non-OOP fix?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NifTK/NiftyNet/pull/380?email_source=notifications&email_token=AEHHUWQFHTNIZ34TQZ4LNY3P2JACTA5CNFSM4HTYIG5KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXTTMRY#issuecomment-501691975, or mute the thread https://github.com/notifications/unsubscribe-auth/AEHHUWURWJ2FSKCQ7QDPKQDP2JACTANCNFSM4HTYIG5A .

NifTK / NiftyNet