Up to higher level directory | |||
Name | Date | Size | |
---|---|---|---|
__init__.py | 21-Aug-2018 | 3K | |
BUILD | 21-Aug-2018 | 1.2K | |
kernels/ | 21-Aug-2018 | ||
ops/ | 21-Aug-2018 | ||
python/ | 21-Aug-2018 | ||
README.md | 21-Aug-2018 | 1.7K |
1 `tf.contrib.data` API 2 ===================== 3 4 NOTE: The `tf.contrib.data` module has been deprecated. Use `tf.data` instead. 5 We are continuing to support existing code using the `tf.contrib.data` APIs in 6 the current version of TensorFlow, but will eventually remove support. The 7 `tf.data` APIs are subject to backwards compatibility guarantees. 8 9 Porting your code to `tf.data` 10 ------------------------------ 11 12 The `tf.contrib.data.Dataset` class has been renamed to `tf.data.Dataset`, and 13 the `tf.contrib.data.Iterator` class has been renamed to `tf.data.Iterator`. 14 Most code can be ported by removing `.contrib` from the names of the classes. 15 However, there are some small differences, which are outlined below. 16 17 The arguments accepted by the `Dataset.map()` transformation have changed: 18 19 * `dataset.map(..., num_threads=T)` is now `dataset.map(num_parallel_calls=T)`. 20 * `dataset.map(..., output_buffer_size=B)` is now 21 `dataset.map(...).prefetch(B)`. 22 23 Some transformations have been removed from `tf.data.Dataset`, and you must 24 instead apply them using `Dataset.apply()` transformation. The full list of 25 changes is as follows: 26 27 * `dataset.dense_to_sparse_batch(...)` is now 28 `dataset.apply(tf.contrib.data.dense_to_sparse_batch(...)`. 29 * `dataset.enumerate(...)` is now 30 `dataset.apply(tf.contrib.data.enumerate_dataset(...))`. 31 * `dataset.group_by_window(...)` is now 32 `dataset.apply(tf.contrib.data.group_by_window(...))`. 33 * `dataset.ignore_errors()` is now 34 `dataset.apply(tf.contrib.data.ignore_errors())`. 35 * `dataset.unbatch()` is now `dataset.apply(tf.contrib.data.unbatch())`. 36 37 The `Dataset.make_dataset_resource()` and `Iterator.dispose_op()` methods have 38 been removed from the API. Please open a GitHub issue if you have a need for 39 either of these. 40