13 lines
505 B
Markdown
13 lines
505 B
Markdown
# basics of batches
|
|
|
|
DataSet and DataLoader classes for batches of data instead of laoding the entire dataset at once
|
|
for training we loop over epochs and batches
|
|
|
|
epoch = 1 forward and backward pass of ALL training samples
|
|
batch_size = number of sample in one forward/backward pass
|
|
number of iterations = number of passes, each pass using [batch_size] number of samples
|
|
-> each epoch gets split up into [number of iterations] passes
|
|
|
|
e.g. 100 samples, batch_size=20 -> 100/20 = 5 iterations for 1 epoch
|
|
|