Feature request - specify python base image

[luarss]

I am building my image using the command: cog build -t model:latest --use-cog-base-image=false --separate-weights but I face an error while parsing. Would it be a good feature to consider for slimmer images?

ⅹ invalid Python version format: strconv.Atoi: parsing "11-slim": invalid syntax

Sample cog.yaml I am using:

build:
  gpu: true
  system_packages:
    - "ffmpeg"
  python_version: "3.11-slim"
  python_packages:
    - "torch==2.3.0"
predict: "predict.py:Predictor"

Other info

• Cog 0.9.20