Closed JeremyCJM closed 1 year ago
Hi Mingyuan,
What is "opt.times" for? I see it is multiplied on the real dataset length in len() method, why do this? Why not just increase the epoch number?
Thanks, Jeremy
Initializing dataloader costs much time because it needs to build datasets for each worker. Therefore I just duplicate the dataset several times to reduce the number of dataloader initialization.
However, the usage of persistent_worker
in the dataloader can significantly relief this situation. opt.times
is then not necessary for current code.
I see, thanks.
Hi Mingyuan,
What is "opt.times" for? I see it is multiplied on the real dataset length in len() method, why do this? Why not just increase the epoch number?
Thanks, Jeremy