Skip to content

Fixed the pooling approach #182

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

Closed
taradepan wants to merge 1 commit into from
Closed

Fixed the pooling approach #182

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion rust/src/embeddings/local/gemma3.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -134,10 +134,12 @@ impl Gemma3Embed for Gemma3Embedder {
self.model.write().unwrap().clear_kv_cache();

// Convert attention_mask to the expected format for pooling
// For decoder-only models like Gemma3, we use LastToken pooling instead of Mean pooling
// The last token contains the most meaningful representation for the entire sequence
let attention_mask = PooledOutputType::from(attention_mask);
let attention_mask = Some(&attention_mask);
let model_output = ModelOutput::Tensor(embeddings.clone());
let pooled_output = Pooling::Mean
let pooled_output = Pooling::LastToken
.pool(&model_output, attention_mask)
.unwrap();
let pooled_output = pooled_output.to_tensor()?;
Expand Down
Loading