NEWS.md

In rstudio/tfdatasets: Interface to 'TensorFlow' Datasets

tfdatasets (development version)

tfdatasets 2.17.0

• Updates for TensorFlow v2.17.0, Keras 3.

tfdatasets 2.9.0

• New dataset_unbatch()
• New dataset_group_by_window()
• New dataset_take_while()
• New as_tensor() and as.array() methods which can be used on TF Datasets with a single element.

tfdatasets 2.7.0

• Added compatability with Tensorflow version 2.7
• as_iterator(), iter_next() and iterate() are is now reexported from {reticualte}.
• New as_array_iterator(), for converting a dataset into an iterable that yields R arrays. (as_iterator() yields tensorflow tensors)
• New dataset_bucket_by_sequence_length()
• New dataset_rejection_resample()
• New dataset_unique()
• New choose_from_datasets()
• sample_from_datasets() gains argument stop_on_empty_dataset.
• dataset_batch() gains arguments num_parallel_calls and deterministic.
• dataset_padded_batch(): Fixed error raised when drop_remainder=TRUE with recent TF versions. Added examples, docs, and tests.
• dataset_concatenate() gains ... and the ability to combine multiple datasets in one call.

tfdatasets 2.6.0

• New dataset_options() for setting and getting dataset options.
• New length() method for tensorflow datasets.
• New dataset_enumerate().
• New random_integer_dataset().
• New dataset_scan(), a stateful variant of dataset_map().
• New dataset_snapshot() for persisting the output of a dataset to disk.
• range_dataset() gains a dtype argument.
• dataset_prefetch() argument buffer_size is now optional, defaults to tf$data$AUTOTUNE

tfdatasets 2.4.0

• Fixed problem when saving models with feature specs (#82).

tfdatasets 1.13.1

• Add datatset_window method.
• Allow purrr style lambda functions in dataset_map.
• Added a NEWS.md file to track changes to the package.
• Added a new feature spec interface that can be used to easily create feature_columns. (#42)

rstudio/tfdatasets documentation built on July 22, 2024, 12:41 a.m.