def _setup_test_model(self):
inputs = input_variable(shape=(1, ), dtype=np.float32)
outputs = input_variable(shape=(1, ), dtype=np.float32)
q = Dense(1, activation=None)(inputs)
loss = squared_error(q, outputs)
return {'inputs': inputs, 'outputs': outputs, 'f': q, 'loss': loss}
def _setup_test_model(self):
inputs = input_variable(shape=(1,), dtype=np.float32)
outputs = input_variable(shape=(1,), dtype=np.float32)
q = Dense(1, activation=None)(inputs)
loss = squared_error(q, outputs)
return {
'inputs': inputs,
'outputs': outputs,
'f': q,
'loss': loss
}
label = input((output_dim), np.float32)
"""
Create model, reader and map
"""
netout = create_model(input_dim, output_dim, hidden_dim, feature)
training_reader = create_reader(data_file_path, True, input_dim, output_dim)
input_map = {
label : training_reader.streams.labels,
feature : training_reader.streams.features
}
"""
Set loss and evaluation functions
"""
loss = squared_error(netout, label)
evaluation = squared_error(netout, label)
lr_per_minibatch=learning_rate_schedule(learning_rate, UnitType.minibatch)
"""
Instantiate the trainer object to drive the model training
See: https://www.cntk.ai/pythondocs/cntk.learners.html
"""
learner = sgd(netout.parameters, lr=lr_per_minibatch)
# Other learners to try
#learner = momentum_sgd(netout.parameters, lr=lr_per_minibatch, momentum = momentum_schedule(0.9))
#learner = adagrad(netout.parameters, lr=lr_per_minibatch)
progress_printer = ProgressPrinter(minibatch_size)
print(str_out)
assert False
results = re.findall("Completed successfully.", str_out)
if len(results) != 2:
print(str_out)
assert False
if __name__=='__main__':
in1 = C.input_variable(shape=1)
labels = C.input_variable(shape=1)
p1 = parameter(shape=1)
p2 = parameter(shape=1)
n = plus(in1, p1, name='n')
z = plus(n, p2, name='z')
ce = squared_error(z, labels)
momentum_schedule = C.momentum_schedule_per_sample(0.9990913221888589)
lr_per_sample = C.learning_parameter_schedule_per_sample(0.007)
dist_learners = [
C.distributed.data_parallel_distributed_learner(C.momentum_sgd([p1], lr_per_sample, momentum_schedule, True)),
C.distributed.data_parallel_distributed_learner(C.momentum_sgd([p2], lr_per_sample, momentum_schedule, True))
]
trainer = C.Trainer(z, ce, dist_learners)
in1_value = [[1]]
label_value = [[0]]
arguments = {in1: in1_value, labels: label_value}
z_output = z.output
def check_samples(learners, expected_number_of_samples):
assert False
results = re.findall("Completed successfully.", str_out)
if len(results) != 2:
print(str_out)
assert False
if __name__ == '__main__':
in1 = C.input_variable(shape=1)
labels = C.input_variable(shape=1)
p1 = parameter(shape=1)
p2 = parameter(shape=1)
n = plus(in1, p1, name='n')
z = plus(n, p2, name='z')
ce = squared_error(z, labels)
momentum_schedule = C.momentum_schedule_per_sample(0.9990913221888589)
lr_per_sample = C.learning_parameter_schedule_per_sample(0.007)
learner1 = C.distributed.data_parallel_distributed_learner(
C.momentum_sgd([p1], lr_per_sample, momentum_schedule, True))
learner1.set_as_metric_aggregator()
dist_learners = [
learner1,
C.distributed.data_parallel_distributed_learner(
C.momentum_sgd([p2], lr_per_sample, momentum_schedule, True))
]
trainer = C.Trainer(z, ce, dist_learners)
in1_value = [[1]]
label_value = [[0]]