Get your VLM running in 3 simple steps on Intel CPUs

Vision Language Models (VLMs) can be run efficiently on local hardware like Intel CPUs by using optimization tools. This tutorial demonstrates a three-step process using Optimum Intel and OpenVINO with the SmolVLM model. The first step involves converting the model to the OpenVINO IR format. Next, the model is optimized through quantization, with options for weight-only or static quantization to reduce its size and accelerate inference. Finally, the quantized model is used to run inference, with the option to leverage Intel GPUs for further performance gains.