metadata

language:
  - en
license: llama3
tags:
  - text-generation-inference
  - transformers
  - unsloth
  - llama
  - trl
  - sft
base_model: unsloth/llama-3-8b-Instruct-bnb-4bit
datasets:
  - zjunlp/Mol-Instructions

Uploaded model

Developed by: kevinkawchak
License: llama3
Finetuned from model : unsloth/llama-3-8b-Instruct-bnb-4bit
Finetuned using dataset : zjunlp/Mol-Instructions
Dataset identification: Molecule-oriented Instructions
Dataset function: Description guided molecule design

Cover Image. META LLAMA 3 COMMUNITY LICENSE AGREEMENT. Built with Meta Llama 3.

A 4-bit quantization of Meta-Llama-3-8B-Instruct was used to reduce training memory requirements when fine-tuning on the zjunlp/Mol-Instructions dataset. (1-2) In addition, the minimum LoRA rank value was utilized to reduce the overall size of created models. In specific, the molecule-oriented instructions description guided molecule design was implemented to answer general questions and general biochemistry questions. General questions were answered with high accuracy, while biochemistry related questions returned 'SELFIES' structures but with limited accuracy.

The notebook featured Torch and Hugging Face libraries using the Unsloth llama-3-8b-Instruct-bnb-4bit quantization model. Training loss decreased steadily from 1.97 to 0.73 over 60 steps. Additional testing regarding the appropriate level of compression or hyperparameter adjustments for accurate SELFIES chemical structures outputs is relevant, as shown in the GitHub notebook for research purposes (3). A 16-bit and reduced 4-bit size were uploaded to Hugging Face. (4-5)

References:

@inproceedings{fang2023mol,
author = {Yin Fang and
Xiaozhuan Liang and
Ningyu Zhang and
Kangwei Liu and
Rui Huang and
Zhuo Chen and
Xiaohui Fan and
Huajun Chen},
title = {Mol-Instructions: {A} Large-Scale Biomolecular Instruction Dataset
for Large Language Models},
booktitle = {{ICLR}},
publisher = {OpenReview.net},
year = {2024},
url = {https://openreview.net/pdf?id=Tlsdsb6l9n}}

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