A few years ago I wrote a program using tensorflow v1.8. The program worked perfectly, however after trying run the code in tensorflow v2, the weights stop updating after the first epoch. I think I have done everything correctly, but I still can’t figure out what the problem is. If the program never worked in the first place I think I would have called it quits, however when I try to implement the old architecture it just keeps failing. Can anyone help ? I have tried everything and I have to submit a thesis soon and am quite stressed about not being able to reproduce the network’s earlier performance.
Is there any reason you can’t run it in Tensorflow 1.8 again?
To expand: Use Anaconda or venv to create a virtual environment to run the code? And then create a requirements.txt file so that others can re-create an environment with the same library versions as the ones you used.