Jamie J. Seol comments

Results 9 comments of


Jamie J. Seol

Added a distance method to the Glove class

:+1:

why initialize the first line of embedding matrix as 0? t.zeros(1, self.embedding_size)

Uh, I wanted to set the line on `padding_idx` to be 0, but in the code it initializes the first line even if `padding_idx` is not zero. Would you PR...

why initialize the first line of embedding matrix as 0? t.zeros(1, self.embedding_size)

`padding_idx` denotes, literally, a padding index for a padding word. The padding word should have zero vector (usually), so it should be initialized as so. About your code, I'm not...

why initialize the first line of embedding matrix as 0? t.zeros(1, self.embedding_size)

@GyeongJoHwang Normally, the "unknown word", or UNK token denotes rare word rather than padding, so making it 0 might be a solution but usually we use random vector or mean...

Where is Expectation

That would give more precise loss value, however, I'm afraid that applying this might consume too much time for training. I quickly wrote a sample code of your idea. With...

Where is Expectation

If you have some good idea for implementation, please go ahead for PR. Otherwise, I'll close this issue. Thank you.

@msummerfield Thanks for the detailed feedback! Awesome. Idea of using the 'faster' loss looks meaningful. The main reason I retained all the details is that the overall loss remains mathematically...

Confused by the loss function.

Thank you for the feedback. Can you provide a reduced, reproducible case sample? Like, small dataset and a configuration for it.

adding APIs for app test user.

Got it. Thanks. But actually, I still feel sorry for quite poor quality of my PR, while still not sure how to improve it.