This repository features a DoRA fine-tuned model for tweet sentiment classification trained as part of VK's LLM course.
Model Description
PEFT (Parameter-Efficient Fine-Tuning) and DoRA (Weight-Decomposed Low-Rank Adaptation) are two techniques used in machine learning to efficiently
adapt large pre-trained neural networks to specific tasks without requiring extensive computational resources.
PEFT (Parameter-Efficient Fine-Tuning): PEFT is a technique that focuses on updating only a small subset of the model’s parameters during fine-tuning,
rather than the entire network. This approach reduces computational costs and memory usage, making it feasible to adapt large models to new tasks on devices
with limited resources. By targeting specific layers or parameters that are most relevant for the task, PEFT achieves significant improvements in efficiency
while maintaining performance.
DoRA (Weight-Decomposed Low-Rank Adaptation): DoRA is a technique for efficiently fine-tuning large pre-trained models by decomposing weight updates into
low-rank components. This method separates the magnitude and direction of weight updates, allowing for focused and parameter-efficient adaptations.
By targeting only the most influential components, DoRA reduces computational and memory demands while maintaining model performance.
This approach is particularly useful for adapting models in resource-constrained environments, enabling scalable fine-tuning with minimal resource usage.
In this case, OuteAI/Lite-Oute-1-300M-Instruct is used as the pre-trained model. This model is further fine-tuned using DoRA to classify tweet's sentiment.
Examples
Before fine-tuning
Tweet: QT @user In the original draft of the 7th book, Remus Lupin survived the Battle of Hogwarts. #HappyBirthdayRemusLupin
