How do I install packages with conda and make them available to the default python backend?

[JamesBowerXanda]

I am attempting to extend the nvcr.io/nvidia/tritonserver:24.04-py3 triton server image so that I can install some extra requirements for the default python backend. Generally I am just able to pip install the requirements in the Dockerfile and these packages can be used by the python backend.

This is great but there are a couple of packages I need to install with conda. In the Python backed README is says

Python backend shipped in the NVIDIA GPU Cloud containers uses Python 3.10. Python backend is able to use the libraries that exist in the current Python environment. These libraries can be installed in a virtualenv, conda environment, or the global system Python.

This implies I can use conda to install packages as well. Is there any way to get the default Python backend to see conda installed libraries or is the only way to use the conda-pack method?