Is dockerizing makes sense in my case?

I have a python code that runs on windows. I want to be able to create executable file from python script files and be able to run it on different machines without installing all dependencies.
The dependecies are in mycode.yaml file (see below). I use anaconda to create the environemnt with conda as conda env create -f mycode.yaml. The code also depends on CUDA. Therefore I have to install anaconda and CUDA Toolkit.

I was thinking to dockerize the whole process and was wondering if that’s possible?
I started with Dockerfile (see below) that installs anaconda and cuda, but:

  1. It seems only the first FROM is executed
  2. I cannot find cuda tag for windows in docker hub. All of them are for ubuntu, linux and redhat.
  3. What docker image base for windows I should include as first layer?

FROM continuumio/anaconda3:latest
FROM nvidia/cuda:12.3.1-base-ubuntu20.04


name: MyCode
  - defaults
  - conda-forge
  - python=3.10
  - numpy
  - plyfile
  - tqdm
  - scikit-learn
  - matplotlib 
  - pandas
  - networkx
  - pip
  - pip: 
    - --extra-index-url
    - torch==2.0.1+cu117 
    - open3d
    - PyYAML