A standard model artifact requires a published conda environment for serving. Where is that inference environment documented for deployment?

1. A prediction request's customer identifier
2. A random notebook cell comment only
3. The model's display-name suffix only
4. runtime.yaml