Container Images

You can import existing images from remote Docker registries like Google Artifact Registry, ECR, Docker Hub, Nvidia and more, simply provide the image argument within your @function or @signal decorator containing an Instance of the Image class

Create a Image instance by importing the Image class

from snowcell.workload import Image

image = Image(
    base_image="docker.io/nvidia/cuda:12.3.1-runtime-ubuntu20.04"
)
.add_commands(["apt-get update -y", "apt-get install neovim -y"])
.add_python_packages(["torch", "cuda"])

Add the image instance to your decorator: This will now run the commands you have specified, add the python packages of your choice and also import the base docker image you'd like to use
```
@client.function(name="my task", cpu=1.5, memory=2, image=image)
def predict(**inputs):
    return inputs['x'] + inputs['y']

predict(x=3, y=2)
```