danielewworrall / harmonicConvolutions

Deep Translation and Rotation Equivariance
MIT License
266 stars 48 forks source link

Error trying to run demo #2

Closed bigsnarfdude closed 7 years ago

bigsnarfdude commented 7 years ago
python train.py 0 mnist deep_stable rotated_data

InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'train_phase' with dtype bool [[Node: train_phase = Placeholder[dtype=DT_BOOL, shape=[], _device="/job:localhost/replica:0/task:0/gpu:0"]()]]

Snippet Error found here: https://gist.github.com/bigsnarfdude/a919f70e9afcf427935a26494800a5e8

StephanGarbin commented 7 years ago

Thanks for noting this! The readme stated incorrectly that tf 0.12 was supported. This bug does not show up in 0.11 which is the version we are targeting. Docs have been amended and support for 0.12 is on our list of todos and should be available in the next couple of weeks. Also note: a host of other warnings show up relating to various API changes in 0.12.

bigsnarfdude commented 7 years ago

Resolved by setting up Tensorflow r.0.11 in virtualenv:

sudo apt-get install python-pip python-dev python-virtualenv
virtualenv --system-site-packages ~/tensorflow
export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow-0.11.0-cp27-none-linux_x86_64.whl
sudo pip install --upgrade $TF_BINARY_URL