You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi @davidtellez,
This is a great project to get started with CPC; however, while running this project in my local environment, I found that while most of the code is compatible, some lines of code are not very compatible with tensorflow 2. For example, fit_genererator throws a warning. Also, in the versions after tensorflow 2.1.0, it has been deprecated.
My proposal
I intend to do the following things to update the project for users using tensorflow 2.
1. Make the existing project compatible with tensorflow 2
2. The callbacks function can incentivize time to train the model and therefore try to look for an alternative, and I will also compare the results when we do not use callback.
3. The project uses LeakyRelu as the activation in many layers; since the loss function is binary_crossentropy, I propose that we use ReLU instead.
Please let me know if you are open to the suggestions made. Also, your feedback is appreciated. Once we are on board, I will be sharing the issues that I discussed in detail. I will provide detailed information about the challenges of running the tensorflow 1.4.0 project on the tensorflow 2 environments.
Question: If this task is assigned to me, should I create a PR for the master branch or create a new branch for the updates.
Note: I'm currently focussing on the train_mode.py file.
The text was updated successfully, but these errors were encountered:
Hi @davidtellez,
This is a great project to get started with CPC; however, while running this project in my local environment, I found that while most of the code is compatible, some lines of code are not very compatible with tensorflow 2. For example,
fit_genererator
throws a warning. Also, in the versions after tensorflow 2.1.0, it has been deprecated.My proposal
I intend to do the following things to update the project for users using tensorflow 2.
1. Make the existing project compatible with tensorflow 2
2. The
callbacks
function can incentivize time to train the model and therefore try to look for an alternative, and I will also compare the results when we do not usecallback.
3. The project uses
LeakyRelu
as the activation in many layers; since the loss function is binary_crossentropy, I propose that we useReLU
instead.Please let me know if you are open to the suggestions made. Also, your feedback is appreciated. Once we are on board, I will be sharing the issues that I discussed in detail. I will provide detailed information about the challenges of running the tensorflow 1.4.0 project on the tensorflow 2 environments.
Question: If this task is assigned to me, should I create a PR for the master branch or create a new branch for the updates.
Note: I'm currently focussing on the train_mode.py file.
The text was updated successfully, but these errors were encountered: