Hi! At MosaicML we currently install FFCV in some of our public docker images, but a recent build is failing due to ffcv==0.0.3 installing sklearn rather than scikit-learn. Attaching the error below.
Would it be possible to upgrade this in FFCV's setup.py before the next release?
18:01:30 Collecting ffcv==0.0.3
18:01:30 Downloading ffcv-0.0.3.tar.gz (53 kB)
18:01:30 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 53.6/53.6 kB 9.9 MB/s eta 0:00:00
18:01:30 Preparing metadata (setup.py): started
18:01:30 Preparing metadata (setup.py): finished with status 'done'
18:01:30 Collecting opencv-python<4.6,>=4.5.5.64
18:01:30 Downloading opencv_python-4.5.5.64-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.5 MB)
18:01:31 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.5/60.5 MB 208.7 MB/s eta 0:00:00
18:01:31 Collecting numba<0.56,>=0.55.0
18:01:31 Downloading numba-0.55.2-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (3.4 MB)
18:01:31 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 3.4/3.4 MB 143.3 MB/s eta 0:00:00
18:01:31 Collecting mmsegmentation<0.23,>=0.22.0
18:01:31 Downloading mmsegmentation-0.22.1-py3-none-any.whl (753 kB)
18:01:31 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 753.8/753.8 kB 152.6 MB/s eta 0:00:00
18:01:31 Collecting terminaltables
18:01:31 Downloading terminaltables-3.1.10-py2.py3-none-any.whl (15 kB)
18:01:31 Collecting pytorch_pfn_extras
18:01:31 Downloading pytorch-pfn-extras-0.6.3.tar.gz (134 kB)
18:01:31 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 134.4/134.4 kB 349.4 MB/s eta 0:00:00
18:01:31 Installing build dependencies: started
18:01:33 Installing build dependencies: finished with status 'done'
18:01:33 Getting requirements to build wheel: started
18:01:33 Getting requirements to build wheel: finished with status 'done'
18:01:33 Preparing metadata (pyproject.toml): started
18:01:34 Preparing metadata (pyproject.toml): finished with status 'done'
18:01:34 Collecting fastargs
18:01:34 Downloading fastargs-1.2.0.tar.gz (12 kB)
18:01:34 Preparing metadata (setup.py): started
18:01:34 Preparing metadata (setup.py): finished with status 'done'
18:01:34 Collecting matplotlib
18:01:34 Downloading matplotlib-3.6.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (11.8 MB)
18:01:34 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 11.8/11.8 MB 68.0 MB/s eta 0:00:00
18:01:34 Collecting sklearn
18:01:34 Downloading sklearn-0.0.post1.tar.gz (3.6 kB)
18:01:34 Preparing metadata (setup.py): started
18:01:35 Preparing metadata (setup.py): finished with status 'error'
18:01:35 error: subprocess-exited-with-error
18:01:35
18:01:35 × python setup.py egg_info did not run successfully.
18:01:35 │ exit code: 1
18:01:35 ╰─> [18 lines of output]
18:01:35 The 'sklearn' PyPI package is deprecated, use 'scikit-learn'
18:01:35 rather than 'sklearn' for pip commands.
18:01:35
18:01:35 Here is how to fix this error in the main use cases:
18:01:35 - use 'pip install scikit-learn' rather than 'pip install sklearn'
18:01:35 - replace 'sklearn' by 'scikit-learn' in your pip requirements files
18:01:35 (requirements.txt, setup.py, setup.cfg, Pipfile, etc ...)
18:01:35 - if the 'sklearn' package is used by one of your dependencies,
18:01:35 it would be great if you take some time to track which package uses
18:01:35 'sklearn' instead of 'scikit-learn' and report it to their issue tracker
18:01:35 - as a last resort, set the environment variable
18:01:35 SKLEARN_ALLOW_DEPRECATED_SKLEARN_PACKAGE_INSTALL=True to avoid this error
18:01:35
18:01:35 More information is available at
18:01:35 https://github.com/scikit-learn/sklearn-pypi-package
18:01:35
18:01:35 If the previous advice does not cover your use case, feel free to report it at
18:01:35 https://github.com/scikit-learn/sklearn-pypi-package/issues/new
18:01:35 [end of output]
18:01:35
18:01:35 note: This error originates from a subprocess, and is likely not a problem with pip.
18:01:35 error: metadata-generation-failed
https://github.com/libffcv/ffcv/blob/b444f0fa8c66bb5132af3ad6ec8db70fb94a3825/setup.py#L55
Hi! At MosaicML we currently install FFCV in some of our public docker images, but a recent build is failing due to
ffcv==0.0.3installingsklearnrather thanscikit-learn. Attaching the error below.Would it be possible to upgrade this in FFCV's
setup.pybefore the next release?