This last weekend, I attempted to make an old python ML code base working, and it was very frustrating. The code is from 2018, and it is maintained, but using

This last weekend, I attempted to make an old python ML code base working, and it was very frustrating. The code is from 2018, and it is maintained, but using python3.8+torch1.4. However, one of its dependency is not maintained anymore and it doesn't work with newer python. I ran into so many other issues even with conda. Things could get worse if you want to merge two ML models using different torch and python versions into one workflow. This is broken.