search

dansuh17 / segan-pytorch

SEGAN pytorch implementation https://arxiv.org/abs/1703.09452
GNU General Public License v3.0
106 stars 32 forks source link

why your test data is in your train batch #18

Closed ScarfBoy0903 closed 5 years ago

ScarfBoy0903 commented 5 years ago

thanks for you code, it helps me a lot, and there r some problems when training, I found that your batch size is 200 and the original paper used 400, have you tried other batch size? and your test data is in your train batch, is there any reason for doing this? and why is the hyper-parameter of activation function used in discriminator is "negative_slope = 0.03" (different from original paper)

thank you

dansuh17 commented 5 years ago

What do you mean by 'test data is in the train batch'? Could you refer the code you are mentioning? Also, various hyperparameters that differ from the original paper might have been a result of my testing with various hyperparameter values. You are free to test them with different values yourself, as the values hardcoded here are not guaranteed to give the best results :)