We added LLaMA on ChatPDF

Hi, HuggingFace members.

i wanted to share some advancements in ChatPDF

  1. Integrating LLaMA 7b model: ChatPDF now leverages the capabilities of the LLaMA 7b model, an open source llm by facebook. This integration allows you to use ChatPDF on a sandbox enviroment and take full control of your sensitive data.

  2. Enhancing search with SPLADE sparse vectors: The search function within ChatPDF has been further optimized by incorporating SPLADE sparse vectors from HuggingFace space. These vectors enable faster and more accurate findings. Combine this with 3072 dimension text-embedding-3-large model, the performance is better than ever. We store both dense and sparse vectors on Pinecone hybrid index. Not to mention we use GPT4 128k model to put as much information as possible from your PDF documents.

These advancements collectively contribute to a more effective PDF chat experience on ChatPDF

Let us know your feedback.