using a GPU: dependencies
uvius opened this issue · comments
The conda installation command
conda env create -f environment_master_gpu.yml
refers to a non-existing file environment_master_gpu.yml
.
However, there is only a file environment_master.yml
.
To make use of my GPU, I installed the latest cudatoolkit and cudnn 11.2, but it turns out that Calamari needs libcuda-10.1 instead. Could you make this dependency explicit and ideally include it in the conda installation example? Maybe by providing the missing file.
The version of CUDA depends on the Tensorflow version (currently 2.4.x). I'll check if using the cudatoolkit
in the envirnoment is a solution.