Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

language_to_ib_point incorrectly converts language to binary matrix for encoder weights #35

Open
nathimel opened this issue Feb 20, 2024 · 0 comments

Comments

@nathimel
Copy link
Collaborator

I suspect that ib_encoder_to_point is correct and that if there is a bug in language_to_ib_point then it comes from converting the language to the encoder, i.e. the Speaker weights, in language_to_ib_encoder_decoder.

I’m pretty sure the LiteralSpeaker has to be the problem. Because if we look at line 230 of agent.py we’ll see it initializes a binary matrix, which is not right.

  • I should consider refactoring these agents to be more flexible at initialization and have this be default behavior if a weight matrix isn’t passed in.
  • What we probably want to do is use the regular Speaker class and pass in weights as one of the kwargs, since CommunicativeAgent will then initialize correctly.
  • This will require some rewriting of the relevant functions in rate_distortion
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant