How much to grab from the dataset each step.
a function that returns a new iterator.
Each element yield'ed by the iterator is a constant array [number, tensor].
The tensor's needn't be disposed manually. Each previous tensor will be disposed before the following yield.
const getIterator = datasetReader(50);
for(let epoch = 0; epoch < 1000; epoch++) {
for(const [timeTook, batchCount, batch] of getIterator()) {
// do something with the batch of data.
}
}
Generated using TypeDoc
This function is used to step over the entire dataset by
batchSize
up until the entire dataset has been gone though. This returns a helper function that itself returns an iterator that is to be used each epoch during training.