aidenhuen / capsule-text-classification Goto Github PK

conv_caps = Capsule( num_capsule=10, dim_capsule=16, routings=3, share_weights=True)(conv)
你好，，num_capsule=10 是和要预测的类别的相对应的嘛？就是说我要是二分类的话，num_capsule是要设置成2嘛？

Hello, I have read you code, but i didn't find primary caps. It seems that after feeding the emdding into the Convolution layer, it then be fed into the convcaps directly. Right?

Hello,
I want to ask if this model performs well in large scale text classification datasets. Just like YELP_REVIEWS_FULL or AMAZON_REVIEWS_FULL. Because as far as I know, most experiments were
conducted on small scale datasets.

where can I download the datasets

Hi:
When I change the parameter of Capsule 'share_weights' from True to False , it can not work with the mistake:
TypeError: unsupported operand type(s) for *: 'int' and 'NoneType'.
Can you give me some suggestions. It seems that the program encounter a problem when initialize the weights parameters of self.W.

您好，我把数据集换成了10分类的问题，为啥出现这样

ValueError: Error when checking target: expected dense_1 to have shape (10,) but got array with shape (500,)