Gradio-multimodal-model

This repository demonstrates Gradio web-interface for multimodal model SmolVLM2.

Model use-cases:

• Image description
• Optical character recognition
• Video description

Get started

1. Clone this repository

git clone https://github.com/comprock15/Gradio-multimodal-model.git

2. Go to the folder where you cloned this repository

cd Gradio-multimodal-model

3. Build docker image:

bash docker/build.sh

4. Start docker image:

bash docker/start.sh

If you start you container for the first time, the webpage can be unavailable for some time as models are not downloaded yet. Wait for a bit and refresh the page.

How to configure parameters

You can configure some parameters in docker/.env:

# Container configuration
IMAGE_NAME=smolvlm2-app-image
CONTAINER_NAME=smolvlm2-app
MODEL_VOLUME=smolvlm2-models
CACHE_VOLUME=smolvlm2-cache

And some runtime parameters in docker/start.sh:

# Runtime configuration
DEVICE=${1:-cuda}     # cuda/cpu
MODEL_SIZE=${2:-256M} # 256M/500M
PORT=${3:-7865}       # any free port