we are a newly formed team setting up our data science plattform based on Jupyter notebooks. For machine learning we aim to use Tensorflow with GPU support. The basic setup is done, users have their own accounts for their own notebooks. The downside is if someone is running some ML script using the GPU, other users will accidentally kill the first script. Are there any best practises to achieve a truely shared environment (sharing GPU ressources)?
ignaciohermosillacornejo
commented
6 years ago
Hi,
we are a newly formed team setting up our data science plattform based on Jupyter notebooks. For machine learning we aim to use Tensorflow with GPU support. The basic setup is done, users have their own accounts for their own notebooks. The downside is if someone is running some ML script using the GPU, other users will accidentally kill the first script. Are there any best practises to achieve a truely shared environment (sharing GPU ressources)?