tfdatasets (version 2.18.0)

dataset_rebatch: Rebatch elements from this dataset into batches of specified size.

Description

dataset_rebatch(N) is functionally equivalent to dataset_unbatch() followed by dataset_batch(N), but it performs only one copy operation, making it more efficient.

Usage

dataset_rebatch(dataset, batch_size, drop_remainder = FALSE, name = NULL)

Value

A dataset.

Arguments

dataset

A dataset.

batch_size

An integer or integer vector specifying batch sizes. If a vector, batch sizes cycle through these values in round-robin order.

drop_remainder

(Optional.) Logical. If TRUE, drops the last batch if it contains fewer elements than batch_size. Defaults to FALSE.

name

(Optional.) Name for the operation.

Details

If batch_size is a vector, it cycles through the provided values in a round-robin manner to determine the size of each batch.

See Also

Other dataset methods: dataset_batch(), dataset_cache(), dataset_collect(), dataset_concatenate(), dataset_decode_delim(), dataset_filter(), dataset_interleave(), dataset_map(), dataset_map_and_batch(), dataset_padded_batch(), dataset_prefetch(), dataset_prefetch_to_device(), dataset_reduce(), dataset_repeat(), dataset_shuffle(), dataset_shuffle_and_repeat(), dataset_skip(), dataset_take(), dataset_take_while(), dataset_window()

Examples

if (FALSE) {
ds <- dataset_range(6) %>% dataset_batch(2) %>% dataset_rebatch(3)
ds %>% as_array_iterator() %>% iterate(print)
# [0, 1, 2], [3, 4, 5]

ds <- dataset_range(7) %>% dataset_batch(4) %>% dataset_rebatch(3)
ds %>% as_array_iterator() %>% iterate(print)
# [0, 1, 2], [3, 4, 5], [6]

ds <- dataset_range(7) %>% dataset_batch(2) %>% dataset_rebatch(3, drop_remainder = TRUE)
ds %>% as_array_iterator() %>% iterate(print)
# [0, 1, 2], [3, 4, 5]

ds <- dataset_range(8) %>% dataset_batch(4) %>% dataset_rebatch(c(2, 1, 1))
ds %>% as_array_iterator() %>% iterate(print)
# [0, 1], [2], [3], [4, 5], [6], [7]
}