Jina.ai logo
Introducing Jina Finetuner-image
Finetuner
Release

Introducing Jina Finetuner

Shubham Saboo
Shubham Saboo

What is Finetuner?

Finetuner is the latest addition to the Jina AI ecosystem that allows you to tune the weights of any deep neural network to create efficient embeddings for search tasks. It aligns with Jina AI’s vision to deliver the last mile of performance-tuning for neural search applications.

What can I do with Finetuner?

Finetuner comes with an intuitive user interface that allows you to perform the following tasks:

  • Refine embeddings: It acts as an ML-powered human-in-the-loop tool for refining embeddings to optimize your pre-trained models in neural search applications.

  • Tune performance: Finetuner lets you perform tasks such as siamese/triplet network, interactive labeling, layer trimming, weights freezing, dimensionality reduction, and much more with just a single line of code - finetuner.fit()

  • Run anywhere: Finetuner is framework agnostic and delivers identical API experience on PyTorch, Keras, and PaddlePaddle deep learning backends.

  • Integrate with Jina AI applications: Finetuner is native to the Jina universe and integrates smoothly with the Jina core, reducing the cost of context-switching between experimentation and production.

Finetuner: Behind the Scenes

Finetuner takes in multiple inputs in the form of data (type-agnostic), model (deep neural network), number of results to be fetched (top-k), along with the human feedback to create a fine-tuned neural network that generates optimized embeddings for better search relevance.

To Get Started

© Jina AI 2020-2021. All rights reserved.