grassknoted / cdae4infoextraction Goto Github PK

While running test_model.py for testing the model, all the training images are loaded, which slows everything down.

Please fix the broken image links in the README

There is a small issue with the decoding part of the model due to which the training fails, here it is:

The error is caused because of the layer defined on line 100 of capsulenet.py.

Fixing this error (in a hacky way) leads to another error which is:

In a nutshell, the model expects a (12, 1) vector, but instead a (13, 1) vector is being passed. There are 13 classes split amongst the 3 domains, so the input layer must be changed.

Please fix the broken image links in the README

Cannot read files due to an UTF-8 to ASCII conversion error at load_custom_dataset() function in capsulenet.py