search

ponder-lab / Hybridize-Functions-Refactoring

Refactorings for optimizing imperative TensorFlow clients for greater efficiency.
Eclipse Public License 2.0
0 stars 0 forks source link

Keras model's call() method should be hybridized #305

Closed khatchad closed 8 months ago

khatchad commented 8 months ago

Consider the following model's call() method:

https://github.com/ponder-lab/samples/blob/39f7644391e664244b45c90868c804abad923eb3/tensorflow_eager_execution/tensorflow_eager_execution.py#L16-L21

As of c5f3fc3735a82c93f37d143c47d88cdc907b0be9, we are saying that this method doesn't have a tensor parameter, but it does have one from a dataset iterable.

subject function module relative path parameters tensor parameter hybrid side-effects recursive autograph experimental_autograph_options experimental_follow_type_hints experimental_implements func input_signature jit_compile reduce_retracing refactoring passing precondition status
samples MyModel.__init__ tensorflow_eager_execution tensorflow_eager_execution/tensorflow_eager_execution.py 1 FALSE FALSE                     CONVERT_EAGER_FUNCTION_TO_HYBRID   3
samples MyModel.call tensorflow_eager_execution tensorflow_eager_execution/tensorflow_eager_execution.py 2 FALSE FALSE                     CONVERT_EAGER_FUNCTION_TO_HYBRID   3