"""``TensorFlowModelDataset`` is a dataset implementation which can save and load
TensorFlow models.
"""
from __future__ import annotations
import copy
import tempfile
from pathlib import PurePath, PurePosixPath
from typing import Any
import fsspec
import tensorflow as tf
from kedro.io.core import (
AbstractVersionedDataset,
DatasetError,
Version,
get_filepath_str,
get_protocol_and_path,
)
TEMPORARY_H5_FILE = "tmp_tensorflow_model.h5"
TEMPORARY_KERAS_FILE = "tmp_tensorflow_model.keras"
[docs]
class TensorFlowModelDataset(AbstractVersionedDataset[tf.keras.Model, tf.keras.Model]):
"""``TensorFlowModelDataset`` loads and saves TensorFlow models.
The underlying functionality is supported by, and passes input arguments through to,
TensorFlow 2.X load_model and save_model methods.
Example usage for the
`YAML API <https://kedro.readthedocs.io/en/stable/data/\
data_catalog_yaml_examples.html>`_:
.. code-block:: yaml
tensorflow_model:
type: tensorflow.TensorFlowModelDataset
filepath: data/06_models/tensorflow_model.h5
load_args:
compile: False
save_args:
overwrite: True
include_optimizer: False
credentials: tf_creds
Example usage for the
`Python API <https://kedro.readthedocs.io/en/stable/data/\
advanced_data_catalog_usage.html>`_:
.. code-block:: pycon
>>> from kedro_datasets.tensorflow import TensorFlowModelDataset
>>> import tensorflow as tf
>>> import numpy as np
>>>
>>> dataset = TensorFlowModelDataset(
... filepath=tmp_path / "data/06_models/tensorflow_model.h5"
... )
>>> model = tf.keras.Sequential(
... [tf.keras.layers.Dense(5, input_shape=(3,)), tf.keras.layers.Softmax()]
... )
>>>
>>> # x = tf.random.uniform((10, 3))
>>> # predictions = model.predict(x)
>>>
>>> dataset.save(model)
>>> loaded_model = dataset.load()
"""
DEFAULT_LOAD_ARGS: dict[str, Any] = {}
DEFAULT_SAVE_ARGS: dict[str, Any] = {}
[docs]
def __init__( # noqa: PLR0913
self,
*,
filepath: str,
load_args: dict[str, Any] | None = None,
save_args: dict[str, Any] | None = None,
version: Version | None = None,
credentials: dict[str, Any] | None = None,
fs_args: dict[str, Any] | None = None,
metadata: dict[str, Any] | None = None,
) -> None:
"""Creates a new instance of ``TensorFlowModelDataset``.
Args:
filepath: Filepath in POSIX format to a TensorFlow model directory prefixed with a
protocol like `s3://`. If prefix is not provided `file` protocol (local filesystem)
will be used. The prefix should be any protocol supported by ``fsspec``.
Note: `http(s)` doesn't support versioning.
load_args: TensorFlow options for loading models.
Here you can find all available arguments:
https://www.tensorflow.org/api_docs/python/tf/keras/models/load_model
All defaults are preserved.
save_args: TensorFlow options for saving models.
Here you can find all available arguments:
https://www.tensorflow.org/api_docs/python/tf/keras/models/save_model
All defaults are preserved, except for "save_format", which is set to "tf".
version: If specified, should be an instance of
``kedro.io.core.Version``. If its ``load`` attribute is
None, the latest version will be loaded. If its ``save``
attribute is None, save version will be autogenerated.
credentials: Credentials required to get access to the underlying filesystem.
E.g. for ``GCSFileSystem`` it should look like `{'token': None}`.
fs_args: Extra arguments to pass into underlying filesystem class constructor
(e.g. `{"project": "my-project"}` for ``GCSFileSystem``).
metadata: Any arbitrary metadata.
This is ignored by Kedro, but may be consumed by users or external plugins.
"""
_fs_args = copy.deepcopy(fs_args) or {}
_credentials = copy.deepcopy(credentials) or {}
protocol, path = get_protocol_and_path(filepath, version)
if protocol == "file":
_fs_args.setdefault("auto_mkdir", True)
self._protocol = protocol
self._fs = fsspec.filesystem(self._protocol, **_credentials, **_fs_args)
self.metadata = metadata
super().__init__(
filepath=PurePosixPath(path),
version=version,
exists_function=self._fs.exists,
glob_function=self._fs.glob,
)
self._tmp_prefix = "kedro_tensorflow_tmp" # temp prefix pattern
# Handle default load and save arguments
self._load_args = copy.deepcopy(self.DEFAULT_LOAD_ARGS)
if load_args is not None:
self._load_args.update(load_args)
self._save_args = copy.deepcopy(self.DEFAULT_SAVE_ARGS)
if save_args is not None:
self._save_args.update(save_args)
self._is_h5 = self._save_args.get("save_format") == "h5"
def _load(self) -> tf.keras.Model:
load_path = get_filepath_str(self._get_load_path(), self._protocol)
with tempfile.TemporaryDirectory(prefix=self._tmp_prefix) as tempdir:
if self._is_h5:
path = str(PurePath(tempdir) / TEMPORARY_H5_FILE) # noqa: PLW2901
else:
# We assume .keras
path = str(PurePath(tempdir) / TEMPORARY_KERAS_FILE) # noqa: PLW2901
self._fs.copy(load_path, path)
# Pass the local temporary directory/file path to keras.load_model
device_name = self._load_args.pop("tf_device", None)
if device_name:
with tf.device(device_name):
model = tf.keras.models.load_model(path, **self._load_args)
else:
model = tf.keras.models.load_model(path, **self._load_args)
return model
def _save(self, data: tf.keras.Model) -> None:
save_path = get_filepath_str(self._get_save_path(), self._protocol)
with tempfile.TemporaryDirectory(prefix=self._tmp_prefix) as tempdir:
if self._is_h5:
path = str(PurePath(tempdir) / TEMPORARY_H5_FILE) # noqa: PLW2901
else:
# We assume .keras
path = str(PurePath(tempdir) / TEMPORARY_KERAS_FILE) # noqa: PLW2901
tf.keras.models.save_model(data, path, **self._save_args)
# Use fsspec to take from local tempfile directory/file and
# put in ArbitraryFileSystem
self._fs.copy(path, save_path)
def _exists(self) -> bool:
try:
load_path = get_filepath_str(self._get_load_path(), self._protocol)
except DatasetError:
return False
return self._fs.exists(load_path)
def _describe(self) -> dict[str, Any]:
return {
"filepath": self._filepath,
"protocol": self._protocol,
"load_args": self._load_args,
"save_args": self._save_args,
"version": self._version,
}
def _release(self) -> None:
super()._release()
self._invalidate_cache()
def _invalidate_cache(self) -> None:
"""Invalidate underlying filesystem caches."""
filepath = get_filepath_str(self._filepath, self._protocol)
self._fs.invalidate_cache(filepath)
