def _infer_model_as_iterable(
self, checkpoint_path, predictions, feed_fn, return_dict):
if feed_fn is None:
feed_dicts = itertools.repeat(None)
else:
def _feed_fn():
while True:
yield feed_fn()
feed_dicts = _feed_fn()
try:
for output_batch in graph_actions.run_feeds_iter(
output_dict=predictions,
feed_dicts=feed_dicts,
restore_checkpoint_path=checkpoint_path):
# Unpack batches into individual predictions
if return_dict:
batch_length = list(output_batch.values())[0].shape[0]
for i in range(batch_length):
yield {key: value[i] for key, value in output_batch.items()}
else:
for pred in output_batch['predictions']:
yield pred
except errors.OutOfRangeError:
# We fall out of the above loop naturally if feed_fn raises StopIteration,
# or we catch an OutOfRangeError if we've reached the end of inputs.
logging.info('Reached end of inputs for predict_iter.')
def _infer_model_as_iterable(self, checkpoint_path, predictions, feed_fn,
return_dict):
if feed_fn is None:
feed_dicts = itertools.repeat(None)
else:
def _feed_fn():
while True:
yield feed_fn()
feed_dicts = _feed_fn()
try:
for output_batch in graph_actions.run_feeds_iter(
output_dict=predictions,
feed_dicts=feed_dicts,
restore_checkpoint_path=checkpoint_path):
# Unpack batches into individual predictions
if return_dict:
batch_length = list(output_batch.values())[0].shape[0]
for i in range(batch_length):
yield {
key: value[i]
for key, value in output_batch.items()
}
else:
for pred in output_batch['predictions']:
yield pred
except errors.OutOfRangeError:
# We fall out of the above loop naturally if feed_fn raises StopIteration,
# or we catch an OutOfRangeError if we've reached the end of inputs.
logging.info('Reached end of inputs for predict_iter.')
