search

danieldjohnson / biaxial-rnn-music-composition

A recurrent neural network designed to generate classical music.
BSD 2-Clause "Simplified" License
1.92k stars 380 forks source link

Any way to resume training? #25

Open Domojz opened 7 years ago

Domojz commented 7 years ago

Is it possible to resume training in any way?

MikhailMS commented 7 years ago

@Domojz You should change trainModel function (multi_training.py) in order to resume your training session and you have to store configurations once they are dumped during training, otherwise it won't be possible. All you need to do is to check in trainModel (or somewhere before, for example in main.py) whether or not you have dumped results, and if you do, use pickle to load 'em. The job is pretty straightforward, good luck!