def load_custom_rnn():
root.update()
filename = filedialog.askopenfilename(initialdir = os.getcwd(), title ='Select file', filetypes=[('Numpy Files', '*.npz')])
if filename != "":
try:
model = RNN(dictionary_size)
log_message = model.load_model_parameters(filename)
generate_sonnet1_button['state'] = 'normal'
generate_sonnet2_button['state'] = 'normal'
bigram_flag = False
write_to_log(log_message, logger_index)
current_model.set("Current Model: Custom RNN")
except(FileNotFoundError):
write_to_log("No such file or directory: %s" %filename, logger_index)
return
def load_custom_rnn():
root.update()
filename = filedialog.askopenfilename(initialdir = os.getcwd(), title ='Select file', filetypes=[('Numpy Files', '*.npz')])
if filename != "":
try:
model = RNN(dictionary_size)
log_message = model.load_model_parameters(filename)
generate_sonnet1_button['state'] = 'normal'
generate_sonnet2_button['state'] = 'normal'
bigram_flag = False
write_to_log(log_message, logger_index)
current_model.set("Current Model: Custom RNN")
except(IOError):
write_to_log("No such file or directory: %s" %filename, logger_index)
return
def load_rnn():
iterations = number_iterations.get()
if iterations == 20:
filename = "RNNModel20.npz"
elif iterations == 40:
filename = "RNNModel40.npz"
elif iterations == 60:
filename = "RNNModel60.npz"
elif iterations == 80:
filename = "RNNModel80.npz"
else:
filename = "RNNModel100.npz"
model = RNN(dictionary_size)
log_message = model.load_model_parameters("TrainedModels/" + filename)
generate_sonnet1_button['state'] = 'normal'
generate_sonnet2_button['state'] = 'normal'
generate_sonnet1_image_button['state'] = 'normal'
generate_sonnet2_image_button['state'] = 'normal'
bigram_flag = False
write_to_log(log_message + " iterations=%d" % iterations, logger_index)
current_model.set("Current Model: RNN %d iterations" % iterations)
def load_rnn_bigram():
iterations = number_iterations.get()
if iterations == 20:
filename = "RNNBigram20.npz"
elif iterations == 40:
filename = "RNNBigram40.npz"
elif iterations == 60:
filename = "RNNBigram60.npz"
elif iterations == 80:
filename = "RNNBigram80.npz"
else:
filename = "RNNBigram100.npz"
model = RNN(cantor_dictionary_size)
log_message = model.load_model_parameters("TrainedModels/" + filename)
generate_sonnet1_button['state'] = 'normal'
generate_sonnet2_button['state'] = 'normal'
generate_sonnet1_image_button['state'] = 'normal'
generate_sonnet2_image_button['state'] = 'normal'
bigram_flag = True
write_to_log(log_message + " iterations=%d" %iterations, logger_index)
current_model.set("Current Model: RNN (Bigram) %d iterations" %iterations)
# Load the dictionary
rhyming_model, syllable_count, syllable_pronounciation, rhyming_dictionary = generate_rhyming_dictionary(
training_data)
dictionary, dictionary_size, word_to_index, start_train, end_train, cantor_dictionary, cantor_dictionary_size, cantor_to_index, start_bigram_train, end_bigram_train, starting_bigrams = load_data(
training_data, syllable_count)
# For graph plotting
loss_array = []
num_examples_array = []
# Train the model
# RNN
if args.mode == "RNN":
print("Generating RNN Model...")
model = RNN(dictionary_size)
print("Training RNN Model with %d iterations..." % number_of_iterations)
losses = model.train_with_sgd(start_train,
end_train,
nepoch=number_of_iterations)
for (num_examples, loss) in losses:
num_examples_array.append(num_examples)
loss_array.append(loss)
plt.plot(num_examples_array, loss_array)
plt.title('Losses of Trained RNN Model %d iterations' %
number_of_iterations)
plt.xlabel('number of examples seen')
plt.ylabel('losses')
plt.grid(True)
plt.show()
model.save_model_parameters(save_to_file)
if not args.mode or not args.iteration:
parser.error("Must provide mode and number of iterations")
sys.exit(0)
number_of_iterations = int(args.iteration)
rhyming_model, syllable_count, syllable_pronounciation, rhyming_dictionary = generate_rhyming_dictionary(
training_data)
dictionary, dictionary_size, word_to_index, start_train, end_train, cantor_dictionary, cantor_dictionary_size, cantor_to_index, start_bigram_train, end_bigram_train, starting_bigrams = load_data(
training_data, syllable_count)
# Generate the neural network mode
# RNN
if args.mode == "RNN":
print("Generating RNN Model...")
model = RNN(dictionary_size)
if number_of_iterations == 20:
model.load_model_parameters('TrainedModels/RNNModel20.npz')
elif number_of_iterations == 40:
model.load_model_parameters('TrainedModels/RNNModel40.npz')
elif number_of_iterations == 60:
model.load_model_parameters('TrainedModels/RNNModel60.npz')
elif number_of_iterations == 80:
model.load_model_parameters('TrainedModels/RNNModel80.npz')
else:
model.load_model_parameters('TrainedModels/RNNModel100.npz')
# GRU
elif args.mode == "GRU":
print("Generating GRU Model...")
model = GRU(dictionary_size)
if number_of_iterations == 20:
# Index for log message
global logger_index
logger_index = 0.0
# Check if generate using bigram or not
bigram_flag = False
# Base width and height of the image
base_width = 440
base_height = 440
# Load rhyming dictionary
rhyming_model, syllable_count, syllable_pronounciation, rhyming_dictionary = generate_rhyming_dictionary(
training_data)
dictionary, dictionary_size, word_to_index, start_train, end_train, cantor_dictionary, cantor_dictionary_size, cantor_to_index, start_bigram_train, end_bigram_train, starting_bigrams = load_data(
training_data, syllable_count)
model = RNN(dictionary_size)
# Load the premade RNN model based on the number of iterations
def load_rnn():
iterations = number_iterations.get()
if iterations == 20:
filename = "RNNModel20.npz"
elif iterations == 40:
filename = "RNNModel40.npz"
elif iterations == 60:
filename = "RNNModel60.npz"
elif iterations == 80:
filename = "RNNModel80.npz"
else:
filename = "RNNModel100.npz"
# Must at least provide mode and iteration
if not args.mode or not args.iteration:
parser.error("Must provide mode and number of iterations")
sys.exit(0)
number_of_iterations = int(args.iteration)
rhyming_model, syllable_count, syllable_pronounciation, rhyming_dictionary = generate_rhyming_dictionary(training_data)
dictionary, dictionary_size, word_to_index, start_train, end_train, cantor_dictionary, cantor_dictionary_size, cantor_to_index, start_bigram_train, end_bigram_train, starting_bigrams = load_data(training_data, syllable_count)
# Generate the neural network mode
# RNN
if args.mode == "RNN":
print("Generating RNN Model...")
model = RNN(dictionary_size)
if number_of_iterations == 20:
model.load_model_parameters('TrainedModels/RNNModel20.npz')
elif number_of_iterations == 40:
model.load_model_parameters('TrainedModels/RNNModel40.npz')
elif number_of_iterations == 60:
model.load_model_parameters('TrainedModels/RNNModel60.npz')
elif number_of_iterations == 80:
model.load_model_parameters('TrainedModels/RNNModel80.npz')
else:
model.load_model_parameters('TrainedModels/RNNModel100.npz')
# GRU
elif args.mode == "GRU":
print("Generating GRU Model...")
model = GRU(dictionary_size)
if number_of_iterations == 20: