👋 Hey everyone,
Back with a new video highlighting a super cool feature that we just added into AnythingLLM where you can create a full fine-tuned model from your chats and documents and run that **locally** with Ollama and LMStudio - no privacy issues or vendor lock-in. At AnythingLLM we believe models trained from your data are your models - we don't gatekeep this technology or restrict you to a single provider.
**This is not a local service** If you want to fine-tune a model locally you can export your data directly from AnythingLLM and do this yourself. A locally running no-code trainer video will be coming soon - subscribe to get notified!
Downloading AnythingLLM: https://anythingllm.com/download
Star on Github: https://github.com/Mintplex-Labs/anything-llm
Send me an email: team@mintplexlabs.com
Chapters
0:00 Introduction to no-code fine-tuning with AnythingLLM
0:25 Disclaimers and caveats
0:47 How to download AnythingLLM
1:27 no-code Fine-tuning overview
2:00 Who this video is for
2:37 Gathering good training data with RAG + AnythingLLM
3:44 Reviewing the dataset we will train on
5:28 Fine-tuning vs RAG for the everyday person
7:00 Ordering a fine-tune
8:04 Privacy & Data Handling
8:44 Selecting a base model
9:18 Uploading our dataset
10:32 What happens once you start
10:54 What you get once tuning is done
11:49 Loading our custom fine-tune into Ollama
15:38 Testing our new fine-tune out!
16:43 How to uninstall a custom fine-tune from Ollama
17:30 Loading our custom fine-tune into LMStudio
20:57 Conclusion and Thanks!