This class performs a model export everytime the new model is better than any
existing model.
Args
name
unique name of this Exporter that is going to be used in the
export path.
serving_input_receiver_fn
a function that takes no arguments and returns
a ServingInputReceiver.
event_file_pattern
event file name pattern relative to model_dir. If
None, however, the exporter would not be preemption-safe. To be
preemption-safe, event_file_pattern must be specified.
compare_fn
a function that compares two evaluation results and returns
true if current evaluation result is better. Follows the signature:
Args:
best_eval_result: This is the evaluation result of the best model.
current_eval_result: This is the evaluation result of current
candidate model.
Returns: True if current evaluation result is better; otherwise,
False.
assets_extra
An optional dict specifying how to populate the assets.extra
directory within the exported SavedModel. Each key should give the
destination path (including the filename) relative to the assets.extra
directory. The corresponding value gives the full path of the source
file to be copied. For example, the simple case of copying a single
file without renaming it is specified as {'my_asset_file.txt':
'/path/to/my_asset_file.txt'}.
as_text
whether to write the SavedModel proto in text format. Defaults to
False.
exports_to_keep
Number of exports to keep. Older exports will be
garbage-collected. Defaults to 5. Set to None to disable garbage
collection.
Raises
ValueError
if any argument is invalid.
Attributes
name
Directory name.
A directory name under the export base directory where exports of
this type are written. Should not be None nor empty.
This boolean is True when this is an export in the
end of training. It is False for the intermediate exports during the
training. When passing Exporter to tf.estimator.train_and_evaluateis_the_final_export is always False if TrainSpec.max_steps is
None.
Returns
The string path to the exported directory or None if export is skipped.