README.md

metadata

base_model: unsloth/meta-llama-3.1-8b-bnb-4bit
tags:
  - text-generation-inference
  - transformers
  - unsloth
  - llama
  - trl
license: apache-2.0
language:
  - en
datasets:
  - OsamaBsher/AITA-Reddit-Dataset
pipeline_tag: text-generation

Reddit AITA Comment Generator

This is a light-hearted experiment in the (sort of) domain-specific fine tuning of small language models. The goal is to see if I can get a SLM to generate comments to posts in the AITA subreddit that:

1. Guess correctly (according to the general concensus in the comments) whether the person is an AH or not. Actually, there are four categories if you check out the subreddit...
2. Generate a reply that fits in more with the style and feel of the other comments than the un-fine-tuned model does.
3. Gets some upvotes 😛

• Developed by: Kerassy
• License: apache-2.0
• Finetuned from model : unsloth/meta-llama-3.1-8b-bnb-4bit

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.