Fine-tune Llama 3.2 Vision models on AMD MI300X GPU using Torchtune, achieving 2.3× better accuracy with 11B vs 90B model on chart-based tasks.| ROCm Blogs
