Lightning-AI/pytorch-lightning

Resuming should allow to differentiate what to resume (steps/opti/weights)

Open

#5 339 ouverte le 3 janv. 2021

Voir sur GitHub
 (26 commentaires) (18 réactions) (0 assignés)Python (3 233 forks)batch import
featurehelp wantedpriority: 1

Métriques du dépôt

Stars
 (26 687 stars)
Métriques de merge PR
 (Merge moyen 9j 15h) (3 PRs mergées en 30 j)

Description

Currently it is possible to either resume only the full training state (epoch/global steps / optimizer / scheduler options / and weights), or only the weights.

I would like to be able to switch the optimizer at some point, i.e. skip restoring optimizer/scheduler, but still load the epoch/global steps. I only see a way to do this with hacks at the moment. Any other way? Could this be a feature, to specify in the trainer.init function specifically what to restore/what not to restore?

Guide contributeur