Open lucas2606-rs opened 2 years ago
It is not wrong. Noting that the shape of the output of the dataloader wrapping the class HSIDatasetPair is [b, 2, 2, h, w, c], where the first three denote batch size, the number of sample pairs, and the number of samples in every pair. The aim of line 112 is to shape it into [2b, 2, h, w, c], so it is also equivalent to x.reshape(2b, 2, h, w, c).
ok,i got it.thx for explaination
i dont understand what's line 112 in Trainer.py mean