def _infer_model(self,
x=None, input_fn=None, feed_fn=None,
batch_size=None, axis=None, proba=False):
# Converts inputs into tf.DataFrame / tf.Series.
batch_size = -1 if batch_size is None else batch_size
if x is not None:
input_fn, feed_fn = _get_predict_input_fn(x, batch_size)
checkpoint_path = saver.latest_checkpoint(self._model_dir)
with ops.Graph().as_default() as g:
random_seed.set_random_seed(self._config.tf_random_seed)
contrib_framework.create_global_step(g)
features, _ = input_fn()
predictions = self._get_predict_ops(features)
if not isinstance(predictions, dict):
predictions = {'predictions': predictions}
# TODO(ipolosukhin): Support batching
if feed_fn is None:
return infer(checkpoint_path, predictions)
preds = {}
while True:
try:
feed_dict = feed_fn()
except StopIteration:
break
if feed_dict is None:
break
outputs = infer(checkpoint_path, predictions, feed_dict=feed_dict)
for key in outputs:
if key not in preds:
preds[key] = []
preds[key].append(outputs[key])
for key in preds:
preds[key] = np.concatenate(preds[key], axis=0)
return preds
def _infer_model(self, x=None, input_fn=None, feed_fn=None, batch_size=None):
# Converts inputs into tf.DataFrame / tf.Series.
batch_size = -1 if batch_size is None else batch_size
if x is not None:
input_fn, feed_fn = _get_predict_input_fn(x, None, batch_size)
checkpoint_path = saver.latest_checkpoint(self._model_dir)
with ops.Graph().as_default() as g:
random_seed.set_random_seed(self._config.tf_random_seed)
contrib_framework.create_global_step(g)
features = self._get_features_from_input_fn(input_fn)
predictions = self._get_predict_ops(features)
return_dict = True
if not isinstance(predictions, dict):
predictions, return_dict = {'predictions': predictions}, False
if feed_fn is None:
preds = infer(checkpoint_path, predictions)
else:
preds = {}
def _feed_fn():
while True:
yield feed_fn()
outputs = graph_actions.run_feeds(
output_dict=predictions,
feed_dicts=_feed_fn(),
restore_checkpoint_path=checkpoint_path)
for key in predictions:
preds[key] = np.concatenate(
[output[key] for output in outputs], axis=0)
if return_dict:
return preds
return preds['predictions']
def _infer_model(self,
x=None,
input_fn=None,
feed_fn=None,
batch_size=None):
# Converts inputs into tf.DataFrame / tf.Series.
batch_size = -1 if batch_size is None else batch_size
if x is not None:
input_fn, feed_fn = _get_predict_input_fn(x, None, batch_size)
checkpoint_path = saver.latest_checkpoint(self._model_dir)
with ops.Graph().as_default() as g:
random_seed.set_random_seed(self._config.tf_random_seed)
contrib_framework.create_global_step(g)
features = self._get_features_from_input_fn(input_fn)
predictions = self._get_predict_ops(features)
return_dict = True
if not isinstance(predictions, dict):
predictions, return_dict = {'predictions': predictions}, False
if feed_fn is None:
preds = infer(checkpoint_path, predictions)
else:
preds = {}
while True:
try:
feed_dict = feed_fn()
except StopIteration:
break
if feed_dict is None:
break
outputs = infer(checkpoint_path,
predictions,
feed_dict=feed_dict)
for key in outputs:
if key not in preds:
preds[key] = []
preds[key].append(outputs[key])
for key in preds:
preds[key] = np.concatenate(preds[key], axis=0)
if return_dict:
return preds
return preds['predictions']
def _infer_model_single(self, checkpoint_path, predictions, feed_fn, return_dict):
if feed_fn is None:
preds = graph_actions.infer(checkpoint_path, predictions)
else:
def _feed_fn():
while True:
yield feed_fn()
outputs = graph_actions.run_feeds(
output_dict=predictions, feed_dicts=_feed_fn(), restore_checkpoint_path=checkpoint_path
)
preds = {key: np.concatenate([output[key] for output in outputs], axis=0) for key in predictions}
return preds if return_dict else preds["predictions"]
def _infer_model(self, x, batch_size=None, axis=None, proba=False):
# Converts inputs into tf.DataFrame / tf.Series.
batch_size = -1 if batch_size is None else batch_size
input_fn, feed_fn = _get_predict_input_fn(x, batch_size)
checkpoint_path = saver.latest_checkpoint(self._model_dir)
with ops.Graph().as_default() as g:
random_seed.set_random_seed(self._config.tf_random_seed)
contrib_framework.create_global_step(g)
features, _ = input_fn()
feed_dict = feed_fn() if feed_fn is not None else None
predictions = self._get_predict_ops(features)
if not isinstance(predictions, dict):
predictions = {'predictions': predictions}
# TODO(ipolosukhin): Support batching
return infer(checkpoint_path, predictions, feed_dict=feed_dict)
def _infer_model(self, input_fn, feed_fn=None, outputs=None):
# Check that model has been trained.
checkpoint_path = saver.latest_checkpoint(self._model_dir)
if not checkpoint_path:
raise NotFittedError("Couldn't find trained model at %s." %
self._model_dir)
with ops.Graph().as_default() as g:
random_seed.set_random_seed(self._config.tf_random_seed)
contrib_framework.create_global_step(g)
features = self._get_features_from_input_fn(input_fn)
predictions = self._get_predict_ops(features)
# If predictions is single output - wrap it into dict, and remember to
# return not a dict.
return_dict = True
if not isinstance(predictions, dict):
predictions, return_dict = {'predictions': predictions}, False
# Filter what to run predictions on, if outputs provided.
if outputs:
existing_keys = predictions.keys()
predictions = {
key: value
for key, value in predictions.items() if key in outputs
}
if not predictions:
raise ValueError(
'Expected to run at least one output from %s, '
'provided %s.' % (existing_keys, outputs))
if feed_fn is None:
preds = graph_actions.infer(checkpoint_path, predictions)
else:
preds = {}
def _feed_fn():
while True:
yield feed_fn()
outputs = graph_actions.run_feeds(
output_dict=predictions,
feed_dicts=_feed_fn(),
restore_checkpoint_path=checkpoint_path)
for key in predictions:
preds[key] = np.concatenate(
[output[key] for output in outputs], axis=0)
if return_dict:
return preds
return preds['predictions']
def _infer_model_single(
self, checkpoint_path, predictions, feed_fn, return_dict):
if feed_fn is None:
preds = graph_actions.infer(checkpoint_path, predictions)
else:
def _feed_fn():
while True:
yield feed_fn()
outputs = graph_actions.run_feeds(
output_dict=predictions,
feed_dicts=_feed_fn(),
restore_checkpoint_path=checkpoint_path)
preds = {
key: np.concatenate([output[key] for output in outputs], axis=0)
for key in predictions}
return preds if return_dict else preds['predictions']
def _infer_model(self, input_fn, feed_fn=None, outputs=None):
# Check that model has been trained.
checkpoint_path = saver.latest_checkpoint(self._model_dir)
if not checkpoint_path:
raise NotFittedError("Couldn't find trained model at %s."
% self._model_dir)
with ops.Graph().as_default() as g:
random_seed.set_random_seed(self._config.tf_random_seed)
contrib_framework.create_global_step(g)
features = self._get_features_from_input_fn(input_fn)
predictions = self._get_predict_ops(features)
# If predictions is single output - wrap it into dict, and remember to
# return not a dict.
return_dict = True
if not isinstance(predictions, dict):
predictions, return_dict = {'predictions': predictions}, False
# Filter what to run predictions on, if outputs provided.
if outputs:
existing_keys = predictions.keys()
predictions = {
key: value for key, value in predictions.items() if key in outputs
}
if not predictions:
raise ValueError('Expected to run at least one output from %s, '
'provided %s.' % (existing_keys, outputs))
if feed_fn is None:
preds = graph_actions.infer(checkpoint_path, predictions)
else:
preds = {}
def _feed_fn():
while True:
yield feed_fn()
outputs = graph_actions.run_feeds(
output_dict=predictions,
feed_dicts=_feed_fn(),
restore_checkpoint_path=checkpoint_path)
for key in predictions:
preds[key] = np.concatenate(
[output[key] for output in outputs], axis=0)
if return_dict:
return preds
return preds['predictions']