Closed n00mkrad closed 1 year ago
Okay, I found out that Ctrl+C does indeed save a checkpoint.
Please clarify this in the readme, I was really scared I might lose hours of training time.
Closed for now though I still would like an option to save a checkpoint every x iterations.
Closed for now though I still would like an option to save a checkpoint every x iterations.
i'll look into it
it basically already does save checkpoints whenever you pass an epoch but i'll check if it's possible to save it every 500 iterations :)
it basically already does save checkpoints whenever you pass an epoch but i'll check if it's possible to save it every 500 iterations :)
Yeah, only problem is that epochs can be very long depending on the dataset size.
yeah i have the same problem hopefully the original publisher will fix it
Running this as we speak, and it seems to work.
However, there are no checkpoints after almost 3k iterations.
1) Will a checkpoint be saved if I stop the training with Ctrl+C?
2) How can I manually specify the checkpoint frequency?