I install nilmtk from nilmtk channel (conda 23.7.4), that has dependancies in channel cobra-fe conda install nilmtk -c nilmtk -c defaults -c conda-feall on older os x to keep 32-bit compatibility. My working setup is to use jupyter notebook that I have to start from special mojave-compatible environment by pinning libffi=3.3 (python=3.9) The error message I get is

The following specifications were found to be incompatible with your system:
  - feature:/osx-64::__osx==10.14.6=0
  - feature:/osx-64::__unix==0=0
  - feature:|@/osx-64::__osx==10.14.6=0
  - feature:|@/osx-64::__unix==0=0
  - nilmtk -> hmmlearn -> __osx[version='>=10.13|>=10.9']
  - nilmtk -> ipykernel -> __linux
  - nilmtk -> ipython -> __unix

Your installed version is: 10.14.6

Can anybody explain what it means?

I tried in an empty new environment, and I tried to manually install all dependancies.

If I install all dependancies, I can install from .git code repository with ! pip install --no-deps git+.git and the only thing pip reports is pip install git+.git --dry-run --report nilmtk_deps_report.json to downgrade matplotlib from 3.7 to 3.1.3. Installed code runs, but with childish errors like 'function' object has no attribute 'items' or 'numpy.bytes_' object has no attribute 'get'

I cannot install nilmtk=0.4.2 from github zip archive because of non-PEP 440 version.

Update: because I cannot install both through anaconda and zip, I suspected that it has conflict with setuptools. E.g. some suggest to downgrade setuptools below v66 to deal with zip files. So I tried to make work for setuptools as easy as possible and I requested specific version and now I have conda install nilmtk=0.4.3 --no-deps -c nilmtk -c defaults -c conda-fe

  - feature:/osx-64::__osx==10.14.6=0
  - nilmtk=0.4.3 -> hmmlearn -> __osx[version='>=10.13|>=10.9']
  - python=3.9 -> __osx[version='>=10.13']
  - python=3.9 -> libcxx[version='>=14.0.6'] -> __osx[version='>=10.9|>=12.0']

Your installed version is: 10.14.6

I have libcxx 19.1.4 installed, and hmmlearn 0.3.3.