Product release

What’s new in Finetuner 0.6?

New CLIP models and ease of use make fine-tuning vector search tasks a breeze

We’re pleased to announce the launch of Jina AI’s Finetuner 0.6, which is now easier to use than ever.

Fine-tuning is an effective way to improve the performance on neural search tasks. However, handling the infrastructure is non-trivial.

Finetuner makes fine-tuning easier, faster and performant by streamlining the workflow and handling all complexity and infrastructure on the cloud. With Finetuner, you can easily uplift pre-trained models to be more performant and production ready.

Want to see how much better fine-tuning makes things? Check our benchmark:

All metrics evaluation on k@20, trained 5 epochs using Adam optimiser with learning rate of 1e-5. You can see the full benchmark here

New CLIP models

We now support the OpenCLIP family plus the Hugging Face variations like openai/clip-large-patch14, openai/clip-vit-large-patch14–336, and openai/clip-vit-base-patch16. To check the full list (30+), simply call finetuner.describe_models().

Easier to use

We heard your feedback and we saw it was very difficult to use a Finetuner if you were not using Jina. In this release, less is more. You can see how much easier it is to use a fine-tuned model:

Use your fine-tuned model with Jina

There are many ways you can use your finetuned model with Jina. Below we’ve outlined just two.

1. Log in
2. Get your token
3. Get your run
4. Use your artifact

Two ways to use your fine-tuned model with Jina. Click to enlarge

Fine-tune and use your model without Jina

You also cannot use Jina if you need a more lightweight experience that just uses DocArray. Again, it’s as simple as:

1. Log in
2. Fit your model
3. Stream your logs
4. Save your artifact
5. Get your model
6. Encode your data

Get started now

1. Head to the Finetuner website to read tutorials on finetuning text-to-text, image-to-image and text-to-image search.
2. Check out the Genshin Impact finetuner notebook.
3. The first time you call finetuner.login() you’ll be prompted to create a Jina AI account.