Good day! Thanks for creating such a nice project. I would use babyGAN models but unfortunately, i can't run it both in docker and on my machine (ubuntu 20, latest cuda-staff, and I tried different versions of TensorFlow).
I have an attribute error when trying to run _encodeImages.py with default parameters using tensorflow=1.10. Exact message:
encode_images.py", line 33, in main
gpu_devices = tf.config.experimental.list_physical_devices('GPU')
AttributeError: module 'tensorflow' has no attribute 'config'
Also, some versions of TensorFlow (including tensorflow=1.10) looking for libcublas.so.10 which are not part of new cuda 11 and therefore should be installed separately.
It would be nice to have exact dependencies to run such a project. Or at least an example of how to run it on Linux machine (maybe Dockefile?).
Thanks in advance.
deng-yongsheng
commented
3 years ago
Good day! Thanks for creating such a nice project. I would use babyGAN models but unfortunately, i can't run it both in docker and on my machine (ubuntu 20, latest cuda-staff, and I tried different versions of TensorFlow).
I have an attribute error when trying to run _encodeImages.py with default parameters using tensorflow=1.10. Exact message:
If we check https://github.com/tensorflow/docs/tree/r1.10/site/en/api_docs/api_docs/python/tf then we can see that there is no tf.config at version 1.10 which you say is the prerequisite.
Also, some versions of TensorFlow (including tensorflow=1.10) looking for libcublas.so.10 which are not part of new cuda 11 and therefore should be installed separately.
It would be nice to have exact dependencies to run such a project. Or at least an example of how to run it on Linux machine (maybe Dockefile?). Thanks in advance.