Rapidsai
Introduction
The RAPIDS suite of software libraries gives you the freedom to execute end-to-end data science and analytics pipelines entirely on GPUs. It relies on NVIDIA® CUDA® primitives for low-level compute optimization, but exposes that GPU parallelism and high-bandwidth memory speed through user-friendly Python interfaces. For more information, please check: NGC: https://ngc.nvidia.com/catalog/containers/nvidia:rapidsai:rapidsai
Versions
0.12
0.13
0.14
0.15
0.16
0.17
21.06
21.10
Commands
ipython3
ipython3
jupyter
python
python3
python3.8
Module
You can load the modules by:
module load ngc
module load rapidsai
Example job
Warning
Using #!/bin/sh -l
as shebang in the slurm job script will cause the failure of some biocontainer modules. Please use #!/bin/bash
instead.
To run rapidsai on our clusters:
#!/bin/bash
#SBATCH -A myallocation # Allocation name
#SBATCH -t 1:00:00
#SBATCH -N 1
#SBATCH -n 1
#SBATCH --job-name=rapidsai
#SBATCH --mail-type=FAIL,BEGIN,END
#SBATCH --error=%x-%J-%u.err
#SBATCH --output=%x-%J-%u.out
module --force purge
ml ngc rapidsai