It's an embeddings database of Wikipedia abstracts with page view data integrated to enable filtering pages based on popularity in addition to similarity.
The first comment in that article has details on the new model. Not the original author but per their testing they said they paid $70 to encode 1M records. The embeddings are 1536 dimensions, which require a lot of vector storage. The HF hub has open models for 384 dimensions or 768 dimensions that work well for a lot of use cases.
That's up to you. Many don't want to open an account and pay in order to explore what's possible. There are LLMs available on the HF Hub, such as google/flan-t5-xl.
If speed and price are concerns, use the FOSS models available on the Hugging Face Hub: https://hf.co/models. Thousands of models, different sizes and tasks. Download locally and fine-tune, if necessary.
txtai is an alternative approach to this. It builds a FAISS (also supports HNSW) index alongside a SQLite database. It works with sentence-transformers models. For example, this model https://huggingface.co/sentence-transformers/all-MiniLM-L6-v... is 384 dimensions and works great for semantic search.
This concept was more for inference on-demand vs training a model. For example, if there was an API call that had a model call as part of it's workflow. Training even a small model still requires serious compute power.
Great looking project here. Absolutely need a local/FOSS option. There's been a number of open-source libraries for LLMs lately that simply call into paid/closed models via APIs. Not exactly the spirit of open-source.
There's already great local/FOSS options such as FLAN-T5 (https://huggingface.co/google/flan-t5-base). Would be great to see a local model like that trained specifically for chat.
Outperforms GPT-3.5 by 16% with less than 1B parameters. There is even a 220M parameter version that scores well. Interesting model to watch.
Referenced snippet from the abstract:
With Multimodal-CoT, our model under 1 billion parameters outperforms the previous state-of-the-art LLM (GPT-3.5) by 16% (75.17%->91.68%) on the ScienceQA benchmark and even surpasses human performance.