My code follows this structure
def train_epoch(epoch):
# assume it loads 1000 samples
# using 10 samples in batch - so it will have 100 batches
train_loader_long = self.DataLoader(...)
# assume it loads 100 samples
# using 10 samples in batch it will have 10 batches
train_loader_short = self.DataLoader(...something else..)
# this should be 10
min_len = min(len(train_loader_short), len(train_loader_long))
for i in range(min_len):
batch_1 = next(iter(train_loader_long))
batch_2 = next(iter(train_loader_short))
...
After 10 batches I will see all sample of loader_short, but only 10% of the loader_long. My question is:
When I call the train_epoch() function for 100 epochs, will it take 100 times SAME samples for the loader_long or will it randomly pick some samples each time. In the latter case, I would (in probability) see all samples if I use sufficiently many epochs.