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.

Sign up for GitHub

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

Jump to bottom

Multi head uses just one set of Q, K, V? #18

Open

daviduarte opened this issue Jun 10, 2021 · 0 comments

daviduarte commented Jun 10, 2021 •

edited

Loading

In transformer.py, in class MultiHeadedSelfAttention() we have the var declaration:

  self.proj_q = nn.Linear(dim, dim)
  self.proj_k = nn.Linear(dim, dim)
  self.proj_v = nn.Linear(dim, dim)

but wasn't suposed to be Q, K and V an independent trainable matrix per head? E.g. if num_head = 12, wasn't that suposed to be like:

set = []
for i in range(12):
    set.append([nn.Linear(dim, dim), nn.Linear(dim, dim), nn.Linear(dim, dim)])

Regards!

The text was updated successfully, but these errors were encountered:

daviduarte changed the title ~~Multi head uses just one set of Q, K, V matrix?~~ Multi head uses just one set of Q, K, V?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment