Visualize training history from a model

#1

I want to generate accuracy/loss vs epoch graph from a trained model. Is it possible to do so? an example image is attached.

acc
acc.png861×650 33 KB

1 Like
#2

You could plot the accuracy e.g. using tensorboard as described here.

#3

@ptrblck is it possible to generate graph with already trained pytorch model?

#4

If you have saved the accuracy values for each epoch, this should be possible.
Otherwise, you would need to rerun the training and plot the metrics for each epoch.

#5

Here is the code in python to do so:

from keras.callbacks import History
history = model.fit(X_test, y_train, epochs = 40, batch_size = 5, verbose = 1)
accuracy = history.history[“accuracy”]
epochs = range(1, len(accuracy) + 1)
import matplotlib.pyplot as plt
plt.plot(epochs, accuracy)
plt.show()