drift_detection_on_images notebook clarification- P-value bootstrapping

[sofiapvcp]

Hi. Thanks for the library and congrats for the greats work.

In the drift_detection_on_images notebook you mention that the p-value being exactly zero is an artifact from the bootstrapping we use for computing the p-value.. Can you elaborate a bit further? I am facing this occurrence. Is it problematic? I tried to increase the n_perm in the kernel_mmd( ) function, but I am still getting the same result. I guess what I'm asking is if this means my results are unreliable, i.e., that drift did in fact not occur the 0. p value I'm getting is not a true reflection of the data.

Thank you!