|
--- |
|
base_model: unsloth/meta-llama-3.1-8b-bnb-4bit |
|
language: |
|
- en |
|
license: apache-2.0 |
|
tags: |
|
- text-generation-inference |
|
- transformers |
|
- unsloth |
|
- llama |
|
- trl |
|
--- |
|
|
|
# Uploaded model |
|
The unsloth/meta-llama-3.1-8b-bnb-4bit model is fine tuned by 30.000 mp speeches and captions from USA Congress, Senate and House. The system promt is below |
|
|
|
system_prompt = """You are an expert captioning assistant specializing in converting a speech transcript into clear, accurate, and viewer-friendly captions. |
|
|
|
### Instruction: |
|
Caption the speech: {} |
|
|
|
### Input: |
|
{} |
|
|
|
### Response: |
|
Caption of the speech: {}""" |
|
|
|
|
|
The data set is curated using |
|
|
|
Judd, Nicholas, Dan Drinkard, Jeremy Carbaugh, and Lindsay Young. congressional-record: A parser for the Congressional Record. Chicago, IL: 2017. https://github.com/unitedstates/congressional-record |
|
|
|
Text is preprocessed by removing President names, Vice President names, party names, and some cliche phrases such as "I reserve the balance of my time","I yield the floor" etc. |
|
- **Developed by:** mesut |
|
- **License:** apache-2.0 |
|
- **Finetuned from model :** unsloth/meta-llama-3.1-8b-bnb-4bit |
|
|
|
This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library. |
|
|
|
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth) |
|
|
