pyext setup Linux: endless dependency resolution

[macumbista]

Hi, I am following the instructions on this page:

https://github.com/SopiMlab/DeepLearningWithAudio/blob/master/utilities/pyext-setup/setup-linux.md

When I reach installing Magenta into the Conda environment from the current directory using pip, then it becomes a death spiral of dependency resolution.

The pip command starts with tensorflow 4.6 and eventually works its way down to tensorflow 1.13 without ever resolving all the dependencies. It fails on note-seq-0.0.4 with the error "pip._vendor.resolvelib.resolvers.ResolutionTooDeep: 2000000" after running for a few hours.

Did I possibly do something wrong or are the pip config files in need of an update somehow?

Thanks!

[macumbista]

Additional info: Ubuntu Studio 22.04 LTS installed directly rather than upgraded from standard Ubuntu install. libasound2-dev and libjack-jackd2-dev packages installed, pure data installed by default, all real-time audio config also already taken care of by the Ubuntu Studio install.