# Module: tf.compat.v1.train.experimental

<br />

Public API for tf._api.v2.train.experimental namespace

## Classes

[`class DynamicLossScale`](https://www.tensorflow.org/api_docs/python/tf/compat/v1/mixed_precision/DynamicLossScale): Loss scale that dynamically adjusts itself.

[`class FixedLossScale`](https://www.tensorflow.org/api_docs/python/tf/compat/v1/mixed_precision/FixedLossScale): Loss scale with a fixed value.

[`class LossScale`](https://www.tensorflow.org/api_docs/python/tf/compat/v1/mixed_precision/LossScale): Base class for all TF1 loss scales.

[`class MaxShardSizePolicy`](https://www.tensorflow.org/api_docs/python/tf/train/experimental/MaxShardSizePolicy): Policy that splits tensors into shards with a max shard size.

[`class MixedPrecisionLossScaleOptimizer`](https://www.tensorflow.org/api_docs/python/tf/compat/v1/mixed_precision/MixedPrecisionLossScaleOptimizer): An optimizer that applies loss scaling.

[`class PythonState`](https://www.tensorflow.org/api_docs/python/tf/train/experimental/PythonState): A mixin for putting Python state in an object-based checkpoint.

[`class ShardByTaskPolicy`](https://www.tensorflow.org/api_docs/python/tf/train/experimental/ShardByTaskPolicy): Policy that splits tensors into shards based on their device spec task.

[`class ShardableTensor`](https://www.tensorflow.org/api_docs/python/tf/train/experimental/ShardableTensor): Tensor wrapper containing data necessary for sharding.

[`class ShardingCallback`](https://www.tensorflow.org/api_docs/python/tf/train/experimental/ShardingCallback): Checkpoint sharding callback function, along with a text description.

## Functions

[`disable_mixed_precision_graph_rewrite(...)`](https://www.tensorflow.org/api_docs/python/tf/compat/v1/mixed_precision/disable_mixed_precision_graph_rewrite): Disables the mixed precision graph rewrite.

[`enable_mixed_precision_graph_rewrite(...)`](https://www.tensorflow.org/api_docs/python/tf/compat/v1/mixed_precision/enable_mixed_precision_graph_rewrite): Enable mixed precision via a graph rewrite.