Overview

From scratch implementation of Llava

Technical Details

• Framework: PyTorch
• Dataset: Flickr
• Category: Computer Vision

Implementation Details

I implemented the Llava using Pytorch on the flickr8000 dataset.

Visual Instruction Tuning

Datasets

flickr 8000: Link

Frameworks:

Pytorch

Results (on T4 GPU Single)

Training epochs: 5

Train loss: 0.23 Val loss: 0.22

Source Code

📁 GitHub Repository: Llava

View the complete implementation, training scripts, and documentation on GitHub.