Ask HN: How do I train a custom LLM/ChatGPT on my own documents in Dec 2023?

Re:Ask HN: How do I train a custom LLM/ChatGPT on my own doc... by treprinum on Monday December 25, 2023 @12:04PM (#38764186)

LlamaIndex can't do chunk-level metadata, only document-level metadata, so you can't put precise references to where materials the LLM synthesized answers from originated, e.g. HTML anchors. Just write your own RAG with Pinecone and OpenAI APIs directly.

[ Parent | on HN ]

Re:Ask HN: How do I train a custom LLM/ChatGPT on my own doc... by seldo on Monday December 25, 2023 @05:16PM (#38766703)

LlamaIndex lets you attach metadata to Nodes which are basically chunks, although that fact is poorly documented! Will fix.

[ Parent | on HN ]

Re:Ask HN: How do I train a custom LLM/ChatGPT on my own doc... by treprinum on Monday December 25, 2023 @06:33PM (#38767371)

Thanks! Even with a better documentation, document importers don't extract node metadata so one needs to write their own "text and metadata extractor" as well. It's then easier to skip LlamaIndex altogether, or just get inspiration from some re-ranking etc. you guys did.

[ Parent | on HN ]

Re:Ask HN: How do I train a custom LLM/ChatGPT on my own doc... by anjanb on Wednesday January 03, 2024 @04:10AM (#38852110)

can you elaborate please

[ Parent | on HN ]