Skip to content

Respect tokenizer.json padding settings in HuggingFaceEmbedder #36071

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
glebashnik merged 1 commit into from
Mar 10, 2026
+3 −2
Merged

Respect tokenizer.json padding settings in HuggingFaceEmbedder #36071

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
5 changes: 3 additions & 2 deletions model-integration/src/main/java/ai/vespa/embedding/HuggingFaceEmbedder.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@
import java.util.Map;
import java.util.logging.Logger;

import static com.yahoo.language.huggingface.ModelInfo.PaddingStrategy.DO_NOT_PAD;
import static com.yahoo.language.huggingface.ModelInfo.TruncationStrategy.LONGEST_FIRST;

@Beta
Expand Down Expand Up @@ -110,11 +111,11 @@ public HuggingFaceEmbedder(OnnxRuntime onnx, Embedder.Runtime runtime, HuggingFa
prependQuery = embedderConfig.prependQuery();
prependDocument = embedderConfig.prependDocument();
var tokenizerPath = modelHelper.getModelPathResolvingIfNecessary(embedderConfig.tokenizerPathReference());
var info = HuggingFaceTokenizer.getModelInfo(tokenizerPath);
var builder = new HuggingFaceTokenizer.Builder()
.addSpecialTokens(true)
.addDefaultModel(tokenizerPath)
.setPadding(false);
var info = HuggingFaceTokenizer.getModelInfo(tokenizerPath);
.setPadding(info.padding() != DO_NOT_PAD);
log.fine(() -> Text.format("'%s' has info '%s'", tokenizerPath, info));
if (info.maxLength() == -1 || info.truncation() != LONGEST_FIRST) {
// Force truncation to max token vector length accepted by model if tokenizer.json contains no valid truncation configuration
Expand Down
Loading