Ablation study

[zaccharieramzi]

I was reading the paper "ResNet strikes back", and the Table 6 caught my attention.

I wanted to know what would be the easiest way to reproduce the results of such a table.

Would I need to simply write a script that sequentially trains all of these models with these different configs? Or is there an already available script to do just that or even better in parallel on SLURM or equivalent?