[question] why doesn't TPOT store checkpoint?

[neel04]

trafficstars

I understand since TPOT would randomly select pipelines, you could use a pseudo-random generator and set a seed for that and export the information to some file.

I used periodic_checkpoint flag and warm_start=True but even then, TPOT doesn't resume from the generation it left off, rather it just starts from scratch.

any workaround to get this checkpoint?

[JDRomano2]

Hi @neel04 , can you provide a minimal set of code we can use to reproduce the issue?

[neel04]

it is not an issue, per se. it is more of a feature request/question about how we can resume TPOT training even after shutting down our computer. apparently, it's not possible since the progress would be stored in the memory which would lose all its information as soon as the computer is switched off.

My question was, could we somehow resume TPOT training even after shutdown?