Currently when using the notebook, and soon the corresponding widget, for running on a whole dataset, we only use 5-8% of the 16 GB GPUs. We should try to see whether it can be optimized. Multiple runner for the data loading? Dask?
Currently when using the notebook, and soon the corresponding widget, for running on a whole dataset, we only use 5-8% of the 16 GB GPUs.
We should try to see whether it can be optimized. Multiple runner for the data loading? Dask?