textsum-gan
textsum-gan copied to clipboard
time for pretraining step
It is mentioned in the repo that the pretraining step should run for some time, please mention after how much time i should interrupt it.
Also i can't use the pretrained npz file as i'm planning to train with my custom dataset.
Actually it should be pretrain for several epochs (e.g., 2 ~ 10), then you can interrupt it.
for initial pretraining step the part of code executed is this, and number of epochs are not mentioned here.
Yes, you could stop running manually when executing several epochs (the epoch count is based on the data size and the batch size). I will fix it later, thanks!