def _get_datasets_and_inputs(outputs):
import tensorflow as tf
all_required_inputs = find_placeholders(outputs)
dataset = tf.get_collection(all_required_inputs[0].name)[0]
inputs = dataset.tensors
_check_the_same(all_required_inputs, inputs)
return dataset, inputs
def _get_arguments_from_loss(loss, optim_method, session, val_outputs,
val_labels, val_method):
import tensorflow as tf
if session is None:
sess = tf.Session()
sess.run(tf.global_variables_initializer())
else:
sess = session
grads_vars = tf.train.GradientDescentOptimizer(0).compute_gradients(
loss)
variables = []
grads = []
for (grad, var) in grads_vars:
if grad is not None:
variables.append(var)
grads.append(grad)
all_required_inputs = _find_placeholders([loss])
dataset = tf.get_collection(all_required_inputs[0].name)[0]
inputs = nest.flatten(dataset._original_tensors)
_check_the_same(all_required_inputs, inputs)
return [
loss, optim_method, sess, dataset, inputs, grads, variables,
loss.graph, val_outputs, val_labels, val_method
]