maulberto3
commented
1 year ago Hi @jalane76, the dataset that SupervisedNE expects is one that is fully compatible with pytorch's DataLoader. Have you tried something like this:
sample = next(iter(torch.utils.data.DataLoader(dataset, batch_size=32)))And it works?
If it doesn't, you might get a recursive error, as per the source code:
def get_minibatch(self) -> Any:
...
try:
batch = next(self.dataloader_iterator)
if batch is None:
self.dataloader_iterator = iter(self.dataloader)
batch = self.get_minibatch()
jalane76
engintoklu
Hello, I'm trying out evotorch to eventually do some black box optimization on the YOLOv8 model. However, I've run into the following error when trying to run SNES on a SupervisedNE problem on the COCO dataset.
Here is the console output:
And here is the code I'm using:
Looking at the source code for get_minibatch in SupervisedNE it appears that if the batch returns
Noneor there is an exception then get_minibatch is called again. This would appear to lead to an infinite regress until overflow and no underlying error is thrown to let me know what the problem is. I tried it on a few other datasets as well and got the same result.Thank you!