Skip to content

some of the investigation types#6

Open
gabar92 wants to merge 6 commits into
linear_predictorfrom
similarity_based_cross_entropy
Open

some of the investigation types#6
gabar92 wants to merge 6 commits into
linear_predictorfrom
similarity_based_cross_entropy

Conversation

@gabar92

@gabar92 gabar92 commented Apr 6, 2022

Copy link
Copy Markdown
Collaborator

No description provided.

Comment thread info_nce.py
# Note: The following code might require a large amount of memory
# in case of large batch size
sim_m = z @ z.T
F.normalize(z, dim=1, p=2)

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ez szerintem nem inplace operátor, így nem csinál semmit

Suggested change
F.normalize(z, dim=1, p=2)
z = F.normalize(z, dim=1, p=2)

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Huh, eddig igy futtattok a simclr-t?

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh semmi, teljesen masik repo. Oke :D

Comment thread info_nce.py
F.normalize(z, dim=1, p=2)
if run_type == "top_k_based_on_cos_sim":
# pairwise similarities between embeddings
cos_sim_m = F.softmax(z @ z.T, dim=1)

@koszpe koszpe May 6, 2022

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

miért van softmax?

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

Successfully merging this pull request may close these issues.

3 participants