Fine tune mistral models on Azure AI services

Question

Hello, I would like to fine tune Mistral models on Azure AI services. But the fine tune option is either not present or not clickable. Is there a roadmap available to know when Azure will allow model fine tuning for mistral models? Thanks, Simon

Answer 1

Hi HOLMES Simon,

Thanks for reaching out to Microsoft Q&A.

Fine-tuning allows adapting a pre-trained model to specific applications or domains. You can fine-tune Mistral 7B on your data to enhance its performance for your specific use case.

Is there a roadmap available to know when Azure will allow model fine tuning for mistral models

Mistral stands out as a Large Language Model with 7.3 billion parameters. It is trained on data that can generate coherent text and perform various natural language processing tasks. Mistral’s premium models will soon be available in Model-as-a-Service (MaaS) through inference APIs and hosted fine-tuning. Specifically, look out for Mistral-7B-V01 and Mistral-7B-Instruct-V01. If you’re interested in fine-tuning Mistral on your own data, some resources available are given below, please spend some time exploring these.

• Brev provides a guide on cost-effective fine-tuning of Mistral 7B. You can explore this tutorial to enhance your natural language processing projects. https://brev.dev/blog/fine-tuning-mistral
• Additionally, there’s a GitHub repository where you can find code for fine-tuning Mistral-7B on specific hardware (such as 3090s or A100s). https://github.com/abacaj/fine-tune-mistral
• Another tutorial walks you through the process of using and fine-tuning Mistral 7B. It covers loading the model, running inference, quantization, merging, and pushing the model to the Hugging Face Hub. https://www.datacamp.com/tutorial/mistral-7b-tutorial

Please 'Upvote'(Thumbs-up) and 'Accept' as an answer if the reply was helpful. This will benefit other community members who face the same issue.