tfm.core.train_utils.try_count_flops

Counts and returns model FLOPs.

tfm.core.train_utils.try_count_flops(
    model: Union[tf.Module, tf.keras.Model],
    inputs_kwargs: Optional[Dict[str, Any]] = None,
    output_path: Optional[str] = None
)

Args
`model`	A model instance.
`inputs_kwargs`	An optional dictionary of argument pairs specifying inputs' shape specifications to getting corresponding concrete function.
`output_path`	A file path to write the profiling results to.

Returns
The model's FLOPs.

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 2024-02-02 UTC.

tfm.core.train_utils.try_count_flops

Args

Returns