Skip to content

feat: Pool embeddings in parallel #53

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

Merged
Pringled merged 2 commits into from
Oct 5, 2026
Merged

feat: Pool embeddings in parallel #53

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
69baaa6
perf: Pool embeddings in parallel and over contiguous rows
Pringled Sep 21, 2026
497d5b6
Trim comments
Pringled Sep 21, 2026
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
30 changes: 18 additions & 12 deletions src/model.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@ use std::{
path::{Path, PathBuf},
sync::Arc,
};
use tokenizers::{Tokenizer, models::ModelWrapper};
use tokenizers::{Tokenizer, models::ModelWrapper, parallelism::MaybeParallelIterator};

/// Static embedding model for Model2Vec.
///
Expand Down Expand Up @@ -399,16 +399,20 @@ impl StaticModel {
.tokenizer
.encode_batch_fast::<String>(truncated.into_iter().map(Into::into).collect(), false)
.expect("tokenization failed");
for encoding in encodings {
let mut token_ids = encoding.get_ids().to_vec();
if let Some(unk_id) = model.unk_token_id {
token_ids.retain(|&id| id as usize != unk_id);
}
if let Some(max_tok) = max_length {
token_ids.truncate(max_tok);
}
embeddings.push(self.pool_ids(token_ids));
}
let pooled: Vec<Vec<f32>> = encodings
.into_maybe_par_iter()
.map(|encoding| {
let mut token_ids = encoding.get_ids().to_vec();
if let Some(unk_id) = model.unk_token_id {
token_ids.retain(|&id| id as usize != unk_id);
}
if let Some(max_tok) = max_length {
token_ids.truncate(max_tok);
}
self.pool_ids(token_ids)
})
.collect();
embeddings.extend(pooled);
}
embeddings
}
Expand Down Expand Up @@ -442,7 +446,9 @@ impl StaticModel {
.unwrap_or(tok);
let scale = model.weights.as_ref().and_then(|w| w.get(tok)).copied().unwrap_or(1.0);
let row = model.embeddings.row(row_idx);
for (s, &v) in sum.iter_mut().zip(row.iter()) {
// Summing over a slice instead of a strided row lets the loop vectorize.
let row = row.as_slice().expect("embedding rows are contiguous");
for (s, &v) in sum.iter_mut().zip(row) {
*s += v * scale;
}
cnt += 1;
Expand Down
Loading