import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
################################################################################
# Graphing Functions #
################################################################################
def plot_learn(values, yAxis, xAxis, title=None, saveTo=None):
"""
Plots the learning curve with train/val for all values. Limited to
7 learning curves on the same graph as we only have 7 colours.
Args:
1. values: Dictionary of tuples of lists, where the tuple is ([train values], [dev values])
and the key is the name of the model for the graph label.
2. yAxis: Either 'Loss', 'F1' or 'Exact Match'
3. xAxis: 'Epochs' or 'Iterations'
4. title: optional title to the graph
5. saveTo: save location for graph
"""
colours = ['b', 'g', 'r', 'c', 'm', 'y', 'k']
for i, (k, (train_values, dev_values)) in enumerate(values.items()):
plt.plot(map(float, train_values), linewidth=2, color=colours[i],
linestyle='--', label="Train {} {}".format(yAxis, k))
if dev_values:
plt.plot(map(float, dev_values), linewidth=2, color=colours[i],
linestyle='-', label="Dev {} {}".format(yAxis, k))
plt.xlabel(xAxis)
plt.ylabel(yAxis)
if title:
plt.title(title)
if yAxis == "Loss":
plt.legend(loc='upper right', shadow=True, prop={'size': 6})
else:
plt.legend(loc='upper left', shadow=True, prop={'size': 6})
assert saveTo
plt.savefig("{}".format(saveTo))
plt.cla()
plt.clf()
plt.close()
def plot_metrics(values, yAxis, xAxis, title=None, saveTo=None):
colours = ['b', 'g', 'r', 'c', 'm', 'y', 'k']
for i, (train_values, dev_values, metric) in enumerate(values):
plt.plot(map(float, train_values), linewidth=2, color=colours[i],
linestyle='-', label="Train {}".format(metric))
if dev_values:
plt.plot(map(float, dev_values), linewidth=2, color=colours[i],
linestyle='--', label="Dev {}".format(metric))
plt.xlabel(xAxis)
plt.ylabel(yAxis)
if title:
plt.title(title)
if yAxis == "Loss":
plt.legend(loc='upper right', shadow=True, prop={'size': 6})
else:
plt.legend(loc='upper left', shadow=True, prop={'size': 6})
assert saveTo
plt.savefig("{}".format(saveTo))
plt.cla()
plt.clf()
plt.close()
