← sources

Best method to train a persona model

source: reddit (reddit:LocalLLaMA)published 2026-10-09 01:05ingested 2026-10-10 02:38

retained text

Hardware available:

2080ti 11gb

5070ti 16gb

Goal:  I've downloaded my whole internet history (reddit posts, emails, chats, etc) and i'd like to train a model to think, act, and talk like me.

I'm currently running a qLORA pass with Qwen3.5-9B as a base, but when i run A/B tests, its unfortunately still pretty clear its not me.

Whats the best method to do this these days?  Obviously would like to stay as local as possible since the corpus contains tons of personal info

Arch document to show the path i've already taken:  [https://markdownpastebin.com/?id=2493bd33b50b48ebbe0a49f437f573a6](https://markdownpastebin.com/?id=2493bd33b50b48ebbe0a49f437f573a6)

document fetched 2026-10-10 00:42 · canonical https://reddit.com/r/LocalLLaMA/comments/1x1835z/best_method_to_train_a_persona_model