net.train() not called again after evaluation finishes

[fawazsammani]

Hi!

From your code, I realized that after you run the eval function and start the next training epoch, you do not turn back to training mode by net.train(), as net.train() is called before the epoch loop. Shouldn't that affect since you are using dropout?

[MIL-VLG]

In eval(), we use the saved model weights to initialize a new net for each call. That is to say, it would not affect the status of the net for training.

[fawazsammani]

Sorry, didn't notice this.