In order to reuse trained models we save training configuration for each experiment in a json file. Together with .pth checkpoints we can generate images. The generate.py script can be seen as a example for embedding WGAN weights/model into other projects.
soumith
commented
5 years ago
In order to reuse trained models we save training configuration for each experiment in a json file. Together with
.pthcheckpoints we can generate images. The generate.py script can be seen as a example for embedding WGAN weights/model into other projects.