Your packages are already large, so I don't think there's much you can do.

• torch: 1.5 GB

• triton: 420 MB

• ray: 170 MB

Using venv not as an isolator but as a package wrapper is a good strategy.

Many AI libraries, especially PyTorch, offer different versions. If you're not going to use a GPU for inference in your container, never install the default version of PyTorch.