InvalidArgumentError: No OpKernel was registered to support Op 'FlowWarp' with these attrs. about flownet2-tf HOT 5 OPEN

Okay. I finally found the issue and resolved the problem. I originally had tensorflow ver 1.4.0 on my linux and I downgraded it to 1.2.0 in order to run this flownet2. My issue was that I still had tensorflow-gpu ver 1.4.0 and after downgrading it to 1.2.0 it worked like nothing happened before.

So you need tensorflow==1.2.0, tensorflow-gpu==1.2.0 and these can be done by

pip install tensorflow==1.2.0
pip install tensorflow-gpu==1.2.0

Hope this resolves your problems as well

from flownet2-tf.

Same problem.

from flownet2-tf.

Same problem.

from flownet2-tf.

I have been tried again in Nvidia-Docker environment.
Using Image: Ubuntu16.04 + CUDA8.0 +CuDNN 5.1
And tensorflow-gpu 1.2.1 + Python2.7

First time I use tensorflow(not gpu version) , it would succeed in make but has this problem.
When I use tensorflow-gpu==1.2.1 and the same environment above, it works fine.

from flownet2-tf.

@ssm9462 works for me. Thanks!

from flownet2-tf.