You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Upon further testing, I found that although the install worked locally when I would install the package with the implemented changes using "pip install git+[repo]" the installation would fail. I closed the PR because of both this issue and the discovery that using "install_requires" in setup script was discouraged as it uses easy_install instead of pip which don't play nicely together. using pyproject.toml seems to be the correct solution to the chicken and egg problem of build tooling as evidenced though discussions linked below: pypa/pip#3691 https://github.com/pandas-dev/pandas/pull/16745/files
I would strongly suggest migrating to this standard.
The text was updated successfully, but these errors were encountered:
In order to allow autosklearn to be built without any unnecessary build dependencies and reduce constraints on users, it would make sense to remove these pre-build dependencies. Pull requests #518 and automl/ConfigSpace#82 attempted to resolve this problem. The following links were the sources of the proposed solution.
https://stackoverflow.com/questions/19919905/how-to-bootstrap-numpy-installation-in-setup-py
numpy/numpy#2434
numpy/numpy#432
Upon further testing, I found that although the install worked locally when I would install the package with the implemented changes using "pip install git+[repo]" the installation would fail. I closed the PR because of both this issue and the discovery that using "install_requires" in setup script was discouraged as it uses easy_install instead of pip which don't play nicely together. using pyproject.toml seems to be the correct solution to the chicken and egg problem of build tooling as evidenced though discussions linked below:
pypa/pip#3691
https://github.com/pandas-dev/pandas/pull/16745/files
I would strongly suggest migrating to this standard.
The text was updated successfully, but these errors were encountered: