Miniconda showing non-sense conflicts

184 Views Asked by At

I'm using Miniconda 4.9.0 on an Ubuntu host. I've set up an environment using python 3.8. Now, I first installed pytorch as follows:

conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

This is working fine. However, next I try to install some more packages:

conda install matplotlib opencv pandas scikit-image scipy tqdm jupyterlab

This results in many conflicts that seem non-sense to me:

Package six conflicts for:
scikit-image -> six[version='>=1.7.3']
scipy -> mkl-service[version='>=2,<3.0a0'] -> six
matplotlib -> cycler[version='>=0.10'] -> six[version='>=1.5']
pandas -> python-dateutil[version='>=2.7.3'] -> six[version='>=1.5']

Package setuptools conflicts for:
python=3.8 -> pip -> setuptools
scikit-image -> matplotlib[version='>=2.0.0'] -> setuptools
matplotlib -> setuptools
jupyterlab -> jinja2[version='>=2.10'] -> setuptools

Package icu conflicts for:
matplotlib -> icu[version='>=58.2,<59.0a0']
opencv -> harfbuzz[version='>=1.5.0,<2.0a0'] -> icu[version='>=58.2,<59.0a0']
scikit-image -> matplotlib[version='>=2.0.0'] -> icu[version='>=58.2,<59.0a0']

Package pytz conflicts for:
scikit-image -> matplotlib[version='>=2.0.0'] -> pytz
matplotlib -> pytz
pandas -> pytz[version='>=2017.2']

Package numpy conflicts for:
opencv -> numpy[version='>=1.9.3,<2.0a0']
opencv -> py-opencv==3.4.2=py37h765d7f9_1 -> numpy[version='>=1.11.3,<2.0a0']

Package jpeg conflicts for:
opencv -> jpeg[version='>=9b,<10a']
scikit-image -> pillow[version='>=4.3.0'] -> jpeg[version='>=9b,<10a']

Package _libgcc_mutex conflicts for:
python=3.8 -> libgcc-ng[version='>=7.3.0'] -> _libgcc_mutex=[build=main]
pandas -> libgcc-ng[version='>=7.3.0'] -> _libgcc_mutex=[build=main]
matplotlib -> libgcc-ng[version='>=7.3.0'] -> _libgcc_mutex=[build=main]
opencv -> libgcc-ng[version='>=7.2.0'] -> _libgcc_mutex=[build=main]
scikit-image -> libgcc-ng[version='>=7.3.0'] -> _libgcc_mutex=[build=main]
scipy -> libgcc-ng[version='>=7.3.0'] -> _libgcc_mutex=[build=main]

Package ssl_match_hostname conflicts for:
jupyterlab -> tornado[version='!=6.0.0,!=6.0.1,!=6.0.2'] -> ssl_match_hostname
matplotlib -> tornado -> ssl_match_hostname

Package tornado conflicts for:
scikit-image -> matplotlib[version='>=2.0.0'] -> tornado
matplotlib -> tornado
jupyterlab -> tornado[version='!=6.0.0,!=6.0.1,!=6.0.2']
jupyterlab -> notebook[version='>=4.3.1'] -> tornado[version='>=4,<6|>=4.1,<7|>=5.0|>=5.0,<7']

Package freetype conflicts for:
opencv -> harfbuzz[version='>=1.5.0,<2.0a0'] -> freetype[version='>=2.9.1,<3.0a0']
opencv -> freetype[version='>=2.8,<2.9.0a0']

For example, the first conflict on six doesn't seem to be a conflict at all, since any version >=2,<3.0a0 matches the dependencies. I have no clue what's wrong here.

Does anybody have a hint?

0

There are 0 best solutions below