Running one of the pre defined experiements raises exceptions

[zwurgl]

Hi, thanks for providing that interesting repo.

When installing it on my Ubuntu 18 with python3 and the following versions of the relevant packages (installing liteflow from github gives me a version 0.0.3 whereas in the dket setup.py a version 0.1.0 is mentioned ...)

$ pip3 freeze | grep -iE '(lite|tensor)' liteflow==0.0.3 tensorboard==1.12.2 tensorboardX==1.2 tensorflow-gpu==1.12.0

I am getting a lot of exceptions when launching one of the stored experiments: python3 ./bin/dket-experiment-run --config experiments/2k-closed.json

Traceback (most recent call last): File "./bin/dket-experiment-run", line 45, in runtime.Experiment.load(CONFIG, logdir=ARGS.logdir, force=ARGS.force).run() File "/usr/local/lib/python3.6/dist-packages/dket-0.0.1-py3.6.egg/dket/runtime.py", line 190, in load [....] File "/usr/local/lib/python3.6/dist-packages/liteflow/utils.py", line 41, in as_scope return next(tf.variable_scope(scope).gen) # pylint: disable=I0011,E1101 AttributeError: 'variable_scope' object has no attribute 'gen'

Any idea where my error could be or what to do about it?

[petrux]

First thing coming on top of my head: you are using a newer TensorFlow version. As specified in the README.md, the best way to go is to run everything in the proper virtual environment. You can get it following the steps:

:~$ git clone git@github.com:dkmfbk/dket.git
:~$ cd dket
:~$ ./bin/dket-venv-setup gpu

the ./bin/dket-venv-setup will install the proper TensorFlow version, which is the 1.2.0---while you are using the 1.12.0.

[zwurgl]

I see! Thanks a lot. Will try!

Best regards!