NNI installation issue with ConfigSpace-0.4.7

[maxxx580]

Describe the issue:

nni[BOHB] has a dependency ConfigSpace==0.4.7, which causes the installation to fail with the error below. The issue is originated from the ConfigSpace package. ConfigSpace uses the numpy in its setup.py without declare it as a setup_requires (source). If numpy is not installed in the current environment, the installation will fail.

Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "/tmp/pip-install-t78p1yeg/configspace_3b56d223c52b4c88851d1ac2a6c0dfa1/setup.py", line 4, in <module>
        import numpy as np
    ModuleNotFoundError: No module named 'numpy'

This problem has been addressed in ConfigSpace==0.4.11 and onwards. I am wondering if NNI could upgrade ConfigSpace to this version or later to avoid the installation error. thanks.

Environment:

• NNI version: 2.2 with BOHB
• Training service (local|remote|pai|aml|etc):local
• Client OS: linux
• Server OS (for remote mode only): linux
• Python version: 3.7
• PyTorch/TensorFlow version: 1.8
• Is conda/virtualenv/venv used?: yes
• Is running in Docker?: no

Configuration:

• Experiment config (remember to remove secrets!): N/A
• Search space: N/A

Log message:

• nnimanager.log: N/A
• dispatcher.log: N/A
• nnictl stdout and stderr: N/A

How to reproduce it?:

1. create an empty environment without numpy
2. run pip install nni[BOHB]

[n810h]

I'm also encountering this. Did you find any workarounds?

[liuzhe-lz]

Fixed in v2.5.