Latentscope

If you've embeded a dataset with nomic-embed-text-v1.5 you can "process SAE" in the embed step.
This will then annotate each row with SAE features from https://enjalot.github.io/latent-taxonomy/articles/about

You can then explore essentially the concepts that the embedding model uses to represent each data point.
You can also filter by a particular SAE feature to see which rows strongly activate for that concept.

<img width="1728" alt="Screenshot 2024-12-20 at 11 05 43 AM" src="https://github.com/user-attachments/assets/1f52ffc3-9ccd-4c6d-89f5-474bdd156dd8" />
<img width="1727" alt="Screenshot 2024-12-20 at 11 05 51 AM" src="https://github.com/user-attachments/assets/c568b020-aa83-47c9-ad3b-597bef6b6533" />
<img width="1725" alt="Screenshot 2024-12-20 at 11 06 19 AM" src="https://github.com/user-attachments/assets/9c49e9a7-c47b-41b2-adfb-7963237b0332" />

Numerous updates to Explore and Setup pages with many contributions from jzhang621

Highlights:
- Explore page redesign with better filtering UX and more screen real estate for the map
- Setup process redesign, step-by-step
- More options for cluster labeling from huggingface, ollama or custom URLs
- Starting to support Sparse Autoencoder features

See closed milestone issues:
https://github.com/enjalot/latent-scope/issues?q=is%3Aissue+milestone%3A0.5+is%3Aclosed

Some more writing on the changes here:
https://enjalot.substack.com/p/hidden-states-and-latent-scope-05

This release fixes a few bugs with the Explore page table UI and nearest neighbor search, making it much more reliable and performant.

Thank you to hydrosquall for issues & PRs! 49 50 52

A new experimental feature for directly visualizing embeddings in the table is ready to try:
<img width="1826" alt="Screenshot 2024-07-26 at 1 38 08 PM" src="https://github.com/user-attachments/assets/dbe89dc4-1c2f-4894-a173-24e972047680">

This release adopts sentence transformers for embedding using local open source models downloaded automatically from HuggingFace hub.

It also keeps track of recently used models and brings it all together in a much improved selector component on the frontend.

<img width="703" alt="Screenshot 2024-07-23 at 3 28 57 PM" src="https://github.com/user-attachments/assets/39b4e4b6-dcaf-42dc-9b7e-465a9c1e9ca5">

Also includes a PR from hydrosquall that fixed a bug using truncated embeddings in the nearest neighbor search.

One minor note: for now truncating of sentence transformers isn't supported as we don't have a way to tell if the model supports it arbitrarily. We could maintain a list of matroyshka enabled models separately.

Export interactive DataMapPlots optionally instead of static thanks to dhruv-anand-aintech

Fixes an unpinned dependency breaking transformers models

Implements 23, creating a UI to easily export static plots using datamapplot
<img width="900" alt=" " src="https://github.com/enjalot/latent-scope/assets/96189/0502ccdf-5eb1-473f-a4db-d122c4996cb5">

Support more filetype inputs thanks to https://github.com/enjalot/latent-scope/pull/40
Support proxy servers / alternate OpenAI compatible endpoints https://github.com/enjalot/latent-scope/pull/44

The requirements.txt has been loosened so Python 3.12 should be enabled and more updated versions of some important pip modules will be installed