CLIP-Vectorizer

This is openAI's CLIP model based API that creates text and image vector-embeddings to be stored and query a vector database.

Steps To run on localhost using Docker

• Make sure Docker is installed and running (and using WSL2 engine if in windows).

• Follow the steps given in Nvidia docs to install nvidia drivers for your distribution (WSL for windows).

• If CUDA drivers are not present or GPU access is not provided to the container, then it will automatically default to computing on CPU.

• This command builds the image to be run inside a container

  docker build -t vectorizer .

• Run the program inside a container using

  docker run -it --gpus all -p 5000:8080 vectorizer

API routes

• /vectors

  Post route for sending text to be embedded in JSON format.
  Example Input JSON:
  {
   "text" : "Your text here",
  }

• /vectors_img

  Post route for sending images to be embedded in form-data format.
  Example Input Form-data:
  Key : file | Value : (Your image file)