def train(cl, dataset):
state = dict(
(key, val)
for key, val in cl.model.state_dict().items()
if key.endswith(".weight"))
test_history, train_history = cl.training(
manager=manager,
nb_epochs=5,
checkpointdir=None,
fold_index=0,
with_validation=False)
train_state = dict(
(key, val)
for key, val in cl.model.state_dict().items()
if key.endswith(".weight"))
for key, val in state.items():
if not np.allclose(val, train_state[key]):
print("--", key)
idx = 0
y_pred_prob, X, y_true, loss, values = cl.testing(
manager=manager,
with_logit=True,
predict=False)
y_pred = np.argmax(y_pred_prob, axis=1)
print(" ** true label : ", y_true[idx])
print(" ** predicted label : ", y_pred[idx])
titles = ["{0}-{1}".format(data.labels[it1], data.labels[it2])
for it1, it2 in zip(y_pred, y_true)]
plot_data(X, labels=titles, nb_samples=5)
plot_history(train_history)
scheduler = lr_scheduler.ReduceLROnPlateau(optimizer=net.optimizer,
mode="min",
factor=0.5,
patience=4,
verbose=True,
min_lr=1e-7)
train_history, valid_history = net.training(
manager=manager,
nb_epochs=(1 if "CI_MODE" in os.environ else 150000),
checkpointdir=None, # outdir,
fold_index=0,
scheduler=scheduler,
with_validation=True)
print(train_history)
print(valid_history)
plot_history(train_history)
#############################################################################
# Testing
# -------
#
# Finaly use the testing set and check the results.
y_pred, X, y_true, loss, values = net.testing(manager=manager,
with_logit=False,
predict=False,
concat_layer_outputs=["flow"])
print(y_pred.shape, X.shape, y_true.shape)
# y_pred = np.expand_dims(y_pred, axis=1)
# data = np.concatenate((y_pred, y_true, X), axis=1)
# plot_data(data, nb_samples=5)
test_history, train_history = cl.training(manager=manager,
nb_epochs=3,
checkpointdir="/tmp/pynet",
fold_index=0,
with_validation=True)
#############################################################################
# You can reload the optimization history at any time and any step.
from pprint import pprint
from pynet.history import History
from pynet.plotting import plot_history
history = History.load("/tmp/pynet/train_0_epoch_2.pkl")
print(history)
plot_history(history)
#############################################################################
# And now predict the labels on the test set.
import numpy as np
from sklearn.metrics import classification_report
from pynet.plotting import plot_data
y_pred, X, y_true, loss, values = cl.testing(manager=manager,
with_logit=True,
predict=True)
pprint(data.labels)
print(classification_report(y_true, y_pred, target_names=data.labels.values()))
titles = [
"{0}-{1}".format(data.labels[it1], data.labels[it2])