def create_RNN_model(args, load_weight_from=None):
''' A wrapper for creating a 'class RNN' instance '''
# Update some dependent args
if hasattr(args, "classes"):
classes = args.classes
elif hasattr(args, "classes_txt"):
classes = lib_io.read_list(args.classes_txt)
else:
raise RuntimeError("The classes are no loaded into the RNN model.")
args.num_classes = len(classes)
args.save_log_to = args.save_model_to + "log.txt"
args.save_fig_to = args.save_model_to + "fig.jpg"
# Create model
device = args.device
model = RNN(args.input_size, args.hidden_size, args.num_layers,
args.num_classes, device).to(device)
model.set_classes(classes)
# Load weights
if load_weight_from:
print(f"Load weights from: {load_weight_from}")
weights = torch.load(load_weight_from)
load_weights(model, weights)
return model
def test_synthesize_audio():
texts = ["hello"]
texts = lib_io.read_list("config/classes_kaggle.names")
for text in texts:
audio = synthesize_audio(text, PRINT=True)
audio.play_audio()
# audio.write_to_file(f"synthesized_audio_{text}.wav")
audio.write_to_file(f"{text}.wav")
def test_synthesize_audio():
texts = ["hello"]
texts = lib_io.read_list("config/classes_kaggle.names")
import os, sys
if not os.path.exists("output/"):
os.makedirs("output/")
for text in texts:
print("=" * 80)
print("Synthesizing " + text + " ...")
audio = synthesize_audio(text, is_print=True)
audio.play_audio()
# audio.write_to_file(f"synthesized_audio_{text}.wav")
audio.write_to_file(f"output/{text}.wav")
def create_RNN_model(args, load_weights_from=None):
''' A wrapper for creating a 'class RNN' instance '''
# Update some dependent args
args.num_classes = len(lib_io.read_list(
args.classes_txt)) # read from "config/classes.names"
args.save_log_to = args.save_model_to + "log.txt"
args.save_fig_to = args.save_model_to + "fig.jpg"
# Create model
device = args.device
model = RNN(args.input_size, args.hidden_size, args.num_layers,
args.num_classes, device).to(device)
# Load weights
if load_weights_from:
print(f"Load weights from: {load_weights_from}")
weights = torch.load(load_weights_from)
load_weights(model, weights)
return model
def setup_classes_labels(load_classes_from, model):
classes = lib_io.read_list(load_classes_from)
print(f"{len(classes)} classes: {classes}")
model.set_classes(classes)