README.md

metadata

language:
  - en
license: llama3
tags:
  - text-generation-inference
  - transformers
  - unsloth
  - llama
  - trl
  - sft
  - not-for-all-audiences
base_model: unsloth/llama-3-8b-bnb-4bit
datasets:
  - mpasila/LimaRP-PIPPA-Mix-8K-Context
  - grimulkan/LimaRP-augmented
  - KaraKaraWitch/PIPPA-ShareGPT-formatted

This is a merge of mpasila/Llama-3-LiPPA-LoRA-8B.

LoRA trained in 4-bit with 8k context using meta-llama/Meta-Llama-3-8B as the base model for 1 epoch.

Dataset used is mpasila/LimaRP-PIPPA-Mix-8K-Context which was made using grimulkan/LimaRP-augmented and KaraKaraWitch/PIPPA-ShareGPT-formatted.

This has been trained on the base model and not the instruct model. The model trained with the instruct model using the same dataset is here: mpasila/Llama-3-Instruct-LiPPA-8B

From quick testing it appears to work fairly well for chatting.

Prompt format: Llama 3 Instruct

Unsloth changed assistant to gpt and user to human.

Uploaded model

• Developed by: mpasila
• License: Llama 3 Community License
• Finetuned from model : unsloth/llama-3-8b-bnb-4bit

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