The tensorflow in the requirements of this project uses 1.x. If a higher version of tensorflow has been installed, you need to modify the code like the following line:
config = tf.compat.v1.ConfigProto()Then install phkit and other modules step by step. Next, I encountered a big trouble. The following sentence was reported as an error. It is not a good way to downgrade tensorflow, because python3.8 does not support tensorflow 1.15.
from tensorflow.contrib.seq2seq import HelperTF1.15 is not supported in python 3.8. If you use conda / docker you could create an environment with python 3.7 which should still support the tensorflow 1.15 version. With conda you can create an environment with: conda create -n tf python=3.7After investigating for a long time, no one provided a solution. I suddenly thought that the Real-Time-Voice-Cloning project no longer uses tensorflow and changed to Torch. I opened the project and found that the two projects are very similar. Perhaps the author of Real-Time-Voice-Cloning is too lazy to transform the code using tensorflow 1.x into 2.x code.
At this time, it was discovered that, unlike the usual project, this project has two levels of directories named zhrtvc. It is better to open the project with the second level of zhrtvc when opening the project, otherwise many of the self-contained packages cannot be imported normally.
Finally, I compromised with the difficulties. After configuring a new virtual machine, I used Python 3.7+tensorflow1.15.2 for the time being.