Fine-Tuning Large Language Models and Non-LLM Models

This repository contains the code and results for fine-tuning various language models (LLMs) and non-large-language models. The models included in this study are:

• distillbert_neural_network
• bert_base_uncased
• bart_large_mnli
• llama3_8b
• mistral_7b
• gemma_7b

Overview

The goal of this project is to compare the performance of different fine-tuned models on a specific task. The accuracies achieved by these models are as follows:

Model	Accuracy (%)
distillbert_neural_network	40.32
bert_base_uncased	66.67
bart_large_mnli	67.30
llama3_8b	74.60
mistral_7b	75.23
gemma_7b	77.11

Results

Dataset Preprocessed

• Dropped sq,sub_topic, sub_sub_topic columns
• Removed all links and emojies
• Replaced Numbers with words
• Droped nan values
• Removed empty rows