Open-source vision-language models (VLMs) have gained traction among machine learning enthusiasts due to their ability to process both text and images for tasks such as image captioning and visual question answering. These models, like the Llama-3.2-11B-Vision-Instruct, are celebrated for their strong zero-shot capabilities, allowing them to perform well in new situations without extra training. However, fine-tuning VLMs presents challenges, including complex tooling, unreliable fine-tuning due to GPU shortages, and costly model serving. Predibase addresses these challenges by simplifying the fine-tuning process through a user-friendly platform that handles data preprocessing and model serving, offering instruction-based fine-tuning for VLMs. They provide a scalable serving infrastructure that allows teams to serve numerous fine-tuned models efficiently. With Predibase, users can format datasets, launch training jobs, and run inferences with ease, as demonstrated by their successful fine-tuning of a Llama-3.2-11B-Vision adapter on a small dataset, achieving significant improvements in accuracy.