dev.langchain4j.model.embedding

Class OnnxEmbeddingModel

java.lang.Object

dev.langchain4j.model.embedding.AbstractInProcessEmbeddingModel

dev.langchain4j.model.embedding.OnnxEmbeddingModel

All Implemented Interfaces:

dev.langchain4j.model.embedding.EmbeddingModel, dev.langchain4j.model.embedding.TokenCountEstimator

public class OnnxEmbeddingModel
extends AbstractInProcessEmbeddingModel

An embedding model that runs within your Java application's process using ONNX runtime.
Many models (e.g., from HuggingFace) can be used, as long as they are in the ONNX format.
Information on how to convert models into ONNX format can be found here.
Many models already converted to ONNX format are available here.

Constructor Summary

Constructors

Constructor and Description
OnnxEmbeddingModel(Path pathToModel) Deprecated. Use OnnxEmbeddingModel(Path, Path, PoolingMode) or OnnxEmbeddingModel(String, String, PoolingMode) instead.
OnnxEmbeddingModel(Path pathToModel, Path pathToTokenizer, PoolingMode poolingMode)
OnnxEmbeddingModel(String pathToModel) Deprecated. Use OnnxEmbeddingModel(Path, Path, PoolingMode) or OnnxEmbeddingModel(String, String, PoolingMode) instead.
OnnxEmbeddingModel(String pathToModel, String pathToTokenizer, PoolingMode poolingMode)

Method Summary

Modifier and Type	Method and Description
protected OnnxBertBiEncoder	model()

Methods inherited from class dev.langchain4j.model.embedding.AbstractInProcessEmbeddingModel

embedAll, estimateTokenCount, loadFromJar

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface dev.langchain4j.model.embedding.EmbeddingModel

embed, embed

Methods inherited from interface dev.langchain4j.model.embedding.TokenCountEstimator

estimateTokenCount, estimateTokenCount

Constructor Detail

OnnxEmbeddingModel

public OnnxEmbeddingModel(Path pathToModel,
                          Path pathToTokenizer,
                          PoolingMode poolingMode)

Parameters:

pathToModel - The path to the modelPath file (e.g., "/path/to/model.onnx")

pathToTokenizer - The path to the tokenizer file (e.g., "/path/to/tokenizer.json")

poolingMode - The pooling model to use. Can be found in the ".../1_Pooling/config.json" file on HuggingFace. Here is an example. "pooling_mode_mean_tokens": true means that PoolingMode.MEAN should be used.

OnnxEmbeddingModel

public OnnxEmbeddingModel(String pathToModel,
                          String pathToTokenizer,
                          PoolingMode poolingMode)

Parameters:

pathToModel - The path to the model file (e.g., "/home/me/model.onnx")

pathToTokenizer - The path to the tokenizer file (e.g., "/path/to/tokenizer.json")

poolingMode - The pooling model to use. Can be found in the ".../1_Pooling/config.json" file on HuggingFace. Here is an example. "pooling_mode_mean_tokens": true means that PoolingMode.MEAN should be used.

OnnxEmbeddingModel

@Deprecated
public OnnxEmbeddingModel(Path pathToModel)

Deprecated. Use OnnxEmbeddingModel(Path, Path, PoolingMode) or OnnxEmbeddingModel(String, String, PoolingMode) instead.

Parameters:

pathToModel - The path to the modelPath file (e.g., "/path/to/model.onnx")

OnnxEmbeddingModel

@Deprecated
public OnnxEmbeddingModel(String pathToModel)

Deprecated. Use OnnxEmbeddingModel(Path, Path, PoolingMode) or OnnxEmbeddingModel(String, String, PoolingMode) instead.

Parameters:

pathToModel - The path to the modelPath file (e.g., "/path/to/model.onnx")

Method Detail

model

protected OnnxBertBiEncoder model()

Specified by:

model in class AbstractInProcessEmbeddingModel