Model Inference

Run inference using ONNX runtime with batch processing support.

Load Model

Load a model for inference:

Copy
-- Load ONNX model
SELECT load_model(
    'model_name',
    '/path/to/model.onnx',
    'onnx'
);

Batch Inference

Run inference on batches for efficiency:

Copy
-- Batch prediction
SELECT id, features,
       model_predict_batch(features, 'model_name') AS predictions
FROM inference_table;

Single Prediction

Copy
-- Single prediction
SELECT model_predict(
    '[1.0, 2.0, 3.0]'::vector,
    'model_name'
) AS prediction;

Model Management

Check loaded models:

Copy
-- List loaded models
SELECT * FROM neurondb.models;

-- Get model info
SELECT model_info('model_name');

Learn More

For detailed documentation on model inference, ONNX runtime, batch processing, and performance optimization, visit:

Model Inference Documentation

Related Topics

• Model Management - Manage model lifecycle
• Embedding Generation - Generate embeddings