# Copyright 2017 Artem Artemev @awav
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import functools
import contextlib
import tensorflow as tf
import pytest
@pytest.fixture
def session_tf():
"""
Session creation pytest fixture.
```
def test_simple(session_tf):
tensor = tf.constant(1.0)
result = session_tf.run(tensor)
# ...
```
In example above the test_simple is wrapped within graph and session created
at `session_tf()` fixture. Session and graph are created per each pytest
function where `session_tf` argument is used.
"""
with session_context() as session:
yield session
def cache_tensor(method):
"""
Caches result for wrapped function wrt default TensorFlow graph.
Whenever function is called under another default graph, execution will be
performed. It does make sense cache tensors: build once, use multiple times
per TensorFlow graph.
Example:
```
@cache_tensor
def create_const():
return tf.constant(1.0, name='wow')
> const1 = create_const()
> const2 = create_const()
> const1 == const2
True
```
"""
cache = {}
@functools.wraps(method)
def wrapper(*args, **kwargs):
graph = tf.get_default_graph()
if graph not in cache:
cache[graph] = method(*args, **kwargs)
return cache[graph]
return wrapper
class session_context(contextlib.ContextDecorator):
def __init__(self, graph=None, close_on_exit=True, **kwargs):
self.graph = graph
self.close_on_exit = close_on_exit
self.session = None
self.session_args = kwargs
def __enter__(self):
graph = tf.Graph() if self.graph is None else self.graph
session = tf.Session(graph=graph, **self.session_args)
self.session = session
session.__enter__()
return session
def __exit__(self, *exc):
session = self.session
session.__exit__(*exc)
if self.close_on_exit:
session.close()
return False
class GPflowTestCase(tf.test.TestCase):
"""
Wrapper for TensorFlow TestCase to avoid massive duplication of resetting
Tensorflow Graph.
"""
_multiprocess_can_split_ = True
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.test_graph = tf.Graph()
@contextlib.contextmanager
def test_context(self, graph=None):
graph = self.test_graph if graph is None else graph
with graph.as_default(), self.test_session(graph=graph) as session:
yield session