Keras + TensorFlow Realtime training chart

Question

I have the following code running inside a Jupyter notebook:

# Visualize training history
from keras.models import Sequential
from keras.layers import Dense
import matplotlib.pyplot as plt
import numpy
# fix random seed for reproducibility
seed = 7
numpy.random.seed(seed)
# load pima indians dataset
dataset = numpy.loadtxt("pima-indians-diabetes.csv", delimiter=",")
# split into input (X) and output (Y) variables
X = dataset[:,0:8]
Y = dataset[:,8]
# create model
model = Sequential()
model.add(Dense(12, input_dim=8, kernel_initializer='uniform', activation='relu'))
model.add(Dense(8, kernel_initializer='uniform', activation='relu'))
model.add(Dense(1, kernel_initializer='uniform', activation='sigmoid'))
# Compile model
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
# Fit the model
history = model.fit(X, Y, validation_split=0.33, epochs=150, batch_size=10, verbose=0)
# list all data in history
print(history.history.keys())
# summarize history for accuracy
plt.plot(history.history['acc'])
plt.plot(history.history['val_acc'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()
# summarize history for loss
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()

The code collects epochs history, then displays the progress history.

---

Q: How can I make the chart change while training so I can see the changes in real time?

Solution 1

There is livelossplot Python package for live training loss plots in Jupyter Notebook for Keras (disclaimer: I am the author).

from livelossplot import PlotLossesKeras

model.fit(X_train, Y_train,
          epochs=10,
          validation_data=(X_test, Y_test),
          callbacks=[PlotLossesKeras()],
          verbose=0)

To see how does it work, look at its source, especially this file: https://github.com/stared/livelossplot/blob/master/livelossplot/outputs/matplotlib_plot.py (from IPython.display import clear_output and clear_output(wait=True)).

A fair disclaimer: it does interfere with Keras output.

Solution 2

Keras comes with a callback for TensorBoard.

You can easily add this behaviour to your model and then just run tensorboard on top of the logging data.

callbacks = [TensorBoard(log_dir='./logs')]
result = model.fit(X, Y, ..., callbacks=callbacks)

And then on your shell:

tensorboard --logdir=/logs

If you need it in your notebook, you can also write your own callback to get metrics while training:

 class LogCallback(Callback):

    def on_epoch_end(self, epoch, logs=None):
        print(logs["train_accuracy"])

This would get the training accuracy at the end of the current epoch and print it. There's some good documentation around it on the official keras site.

Comments

[Shamoon]

Does this only with for Jupyter notebooks?

[Lucas Farias]

But does TensorBoard show you live trainning data or can you only see it after training the model?

[Dinesh]

I had a question after using the library, livelossplot. I observed the following metrics: validation (min: 113.068, max: 1852.994, cur: 117.239). I clearly get the meaning of 'min' and 'max', but what does 'cur' actually mean ?

[Piotr Migdal]

Now it is not only for Jupyter.

[glicerico]

@LucasFarias you can see it real time with Tensorboard

[jtlz2]

@Dinesh "Current", surely?