The dependency of medaka=1.1.2 which fixes the dep for tensorflow=2.2.0 (released in 2019....) produces a lot of constraints to install this software with correct GPU support for tensorflow nowadays.
conda-forge packages for tensorflow support the CUDA builds only for tensorflow>=2.4.0 .
So, if i'm not following it wrong, your current approach for installation with a conda environment is producing a CPU only build.
To overcome this issue, could you consider bumping the medaka dependency used ?
In order to support more recent tensorflows, it should be to at least medaka=1.6.1
Hello,
The dependency of
medaka=1.1.2
which fixes the dep fortensorflow=2.2.0
(released in 2019....) produces a lot of constraints to install this software with correct GPU support for tensorflow nowadays.conda-forge packages for tensorflow support the CUDA builds only for tensorflow>=2.4.0 .
So, if i'm not following it wrong, your current approach for installation with a conda environment is producing a CPU only build.
To overcome this issue, could you consider bumping the medaka dependency used ? In order to support more recent tensorflows, it should be to at least medaka=1.6.1
thanks.