def main(modelfile: str,
features: List[float],
print_results: bool = True) -> List[Dict[str, Any]]:
"""
Evaluate the model described in ``modelfile`` with ``inputvec`` as input
data.
Parameters
----------
features : List[float]
print_results : bool
Print results if True. Always return results.
Returns
-------
List of possible answers, reverse-sorted by probability.
"""
model = utils.get_model(modelfile)
if not model:
return []
x = np.array([features])
model_output = get_model_output(model, x)
results = get_results(model_output, model["outputs"])
if print_results:
show_results(results, n=10)
return results
def main(model_file, test_data, verbose=True):
""" Evaluate a model
Parameters
----------
model_file : string
Path to a model file
test_data : string
Path to a testdata.tar file
Returns
-------
Testing results
"""
model = utils.get_model(model_file)
data = utils.get_data(test_data)
if data is None:
logging.error("Data could not be loaded. Stop testing.")
return
x_vec, y_vec = data
correct = 0
total = 0
for x, y in zip(x_vec, y_vec):
x = numpy.array([x])
y_pred = evaluate.get_model_output(model, x)
y_pred = numpy.argmax(y_pred)
if y_pred == y[0]:
correct += 1
total += 1
if verbose and total % 100 == 0:
print("%i: %0.2f" % (total, float(correct)/total))
print("Correct: %i/%i = %0.2f of total correct" %
(correct, total, float(correct)/total))
return float(correct)/total
def main(model_file: str, test_data: str, verbose=True) -> float:
"""
Evaluate a model
Parameters
----------
model_file : str
Path to a model file
test_data : str
Path to a testdata.hdf5 file
Returns
-------
Testing results
"""
model = utils.get_model(model_file)
data = utils.get_data(test_data)
x_vec, y_vec = data
correct = 0
total = 0
for x, y in zip(x_vec, y_vec):
x = numpy.array([x])
y_pred = evaluate.get_model_output(model, x)
y_pred = numpy.argmax(y_pred)
if y_pred == y[0]:
correct += 1
total += 1
if verbose and total % 100 == 0:
print("%i: %0.2f" % (total, float(correct) / total))
print("Correct: %i/%i = %0.2f of total correct" %
(correct, total, float(correct) / total))
return float(correct) / total
def main(model_file, test_data, verbose=True):
""" Evaluate a model
Parameters
----------
model_file : string
Path to a model file
test_data : string
Path to a testdata.tar file
Returns
-------
Testing results
"""
model = utils.get_model(model_file)
data = utils.get_data(test_data)
if data is None:
logging.error("Data could not be loaded. Stop testing.")
return
x_vec, y_vec = data
correct = 0
total = 0
for x, y in zip(x_vec, y_vec):
x = numpy.array([x])
y_pred = evaluate.get_model_output(model, x)
y_pred = numpy.argmax(y_pred)
if y_pred == y[0]:
correct += 1
total += 1
if verbose and total % 100 == 0:
print("%i: %0.2f" % (total, float(correct) / total))
print("Correct: %i/%i = %0.2f of total correct" %
(correct, total, float(correct) / total))
return float(correct) / total
def main(model_file, model_output_file, training_data, batch_size,
learning_rate, epochs):
"""Train model_file with training_data."""
data = utils.get_data(training_data)
if data is None:
logging.error("Data could not be loaded. Stop training.")
return
x, y = data
assert y is not None
model = utils.get_model(model_file)
minibatch_gradient_descent(model, x, y, batch_size, learning_rate, epochs)
utils.write_model(model, model_output_file)
def main(modelfile, features, print_results=True):
"""Evaluate the model described in ``modelfile`` with ``inputvec`` as
input data.
Parameters
----------
features : list of floats
print_results : bool
Print results if True. Always return results.
Returns
-------
List of possible answers, reverse-sorted by probability.
"""
model = utils.get_model(modelfile)
if not model:
return []
x = numpy.array([features])
model_output = get_model_output(model, x)
results = get_results(model_output, model['outputs'])
if print_results:
show_results(results, n=10)
return results