Working on GPU server for deep learning application via ssh is very challenges. One the challenges is if you want to plot your training accuracy on graph because our server is text based and fully command line. For solving this proble we can save the value of accuracy and loss to the file json. This is the code.
history = model.fit(X, Y, validation_split=0.33, epochs=150, batch_size=10, verbose=0)
print(history.history.keys())
with open(‘file.json’, ‘w’) as f:
json.dump(history.history, f)
After file json are generated, we use web based tools