Sherif Ali Ahmad

Results 9 comments of Sherif Ali Ahmad

Some people are trying to add a [similar feature](https://github.com/pytorch/pytorch/issues/26889) to the Python version.

Reducing thresholds adds red dots to output with the runtime of demo increased a lot.

Do you know a configuration (Number of samples, batch size, epochs, learning rate, etc) that is guaranteed to overfit successfully? I can't increase batch size to 4.

Re-downloading the repository fixed the issue. It seems I broke something.

You can close the issue. Thanks for help.

What is the best training configuration for generalization on test set? This is with batch size of 2: ![Screenshot from 2019-03-24 15-57-07](https://user-images.githubusercontent.com/37023248/54880490-b5241580-4e4d-11e9-9490-b68bfb5917c6.png) Blue: Training. Red: Testing.

I also have the bug. Ubuntu: (64 bit) 18.10. ![atom bug](https://user-images.githubusercontent.com/37023248/53440605-a0e72700-3a0d-11e9-87a5-c79b74f02231.png)

I have: Ubuntu 18.10 (64 bit). All is up to date. CPU: Intel Core i5.