Module: tf.experimental

Public API for tf.experimental namespace.

Modules

dlpack module: Public API for tf.experimental.dlpack namespace.

dtensor module: Public API for tf.experimental.dtensor namespace.

extension_type module: Public API for tf.experimental.extension_type namespace.

numpy module: # tf.experimental.numpy: NumPy API on TensorFlow.

tensorrt module: Public API for tf.experimental.tensorrt namespace.

Classes

class BatchableExtensionType: An ExtensionType that can be batched and unbatched.

class DynamicRaggedShape: The shape of a ragged or dense tensor.

class ExtensionType: Base class for TensorFlow ExtensionType classes.

class ExtensionTypeBatchEncoder: Class used to encode and decode extension type values for batching.

class Optional: Represents a value that may or may not be present.

class RowPartition: Partitioning of a sequence of values into contiguous subsequences ("rows").

class StructuredTensor: A multidimensional collection of structures with the same schema.

Functions

async_clear_error(...): Clear pending operations and error statuses in async execution.

async_scope(...): Context manager for grouping async operations.

dispatch_for_api(...): Decorator that overrides the default implementation for a TensorFlow API.

dispatch_for_binary_elementwise_apis(...): Decorator to override default implementation for binary elementwise APIs.

dispatch_for_binary_elementwise_assert_apis(...): Decorator to override default implementation for binary elementwise assert APIs.

dispatch_for_unary_elementwise_apis(...): Decorator to override default implementation for unary elementwise APIs.

function_executor_type(...): Context manager for setting the executor of eager defined functions.

register_filesystem_plugin(...): Loads a TensorFlow FileSystem plugin.

unregister_dispatch_for(...): Unregisters a function that was registered with @dispatch_for_*.

float8_e4m3fn Instance of tf.dtypes.DType

8-bit float with 4 exponent bits and 3 mantissa bits, with extended finite range. This type has no representation for inf, and only two NaN values: 0xFF for negative NaN, and 0x7F for positive NaN.

float8_e5m2 Instance of tf.dtypes.DType

8-bit float with 5 exponent bits and 2 mantissa bits.