tf.train.experimental.ShardableTensor | TensorFlow v2.16.1 (original) (raw)

tf.train.experimental.ShardableTensor

Stay organized with collections Save and categorize content based on your preferences.

Tensor wrapper containing data necessary for sharding.

View aliases

Compat aliases for migration

tf.compat.v1.train.experimental.ShardableTensor

tf.train.experimental.ShardableTensor(
    _tensor_save_spec: saveable_object.SaveSpec,
    tensor: tf.Tensor,
    dtype: tf.dtypes.DType,
    device: tf.DeviceSpec,
    name: str,
    shape: tf.TensorShape,
    slice_spec: tf.Variable.SaveSliceInfo,
    checkpoint_key: str,
    trackable: base.Trackable
)

The tensor representation used as inputs to pre-made and customtf.train.experiemental.ShardingCallbacks, which can be specified using theexperimental_sharding_callback option in tf.train.CheckpointOptions.

Attributes
tensor	Dataclass field
dtype	Dataclass field
device	Dataclass field
name	Dataclass field
shape	Dataclass field
slice_spec	Dataclass field
checkpoint_key	Dataclass field
trackable	Dataclass field

Methods

`eq`

__eq__(
    other
)

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates. Some content is licensed under the numpy license.

Last updated 2024-04-26 UTC.