def load_data_engin(self, train_addr, valid_addr):
torch.cuda.empty_cache()
self.train_addr = train_addr
self.valid_addr = valid_addr
self.train = Data_Engin(method=self.method,
mono=self.mono,
address=self.train_addr,
spectra_type=self.spectra_type,
device=self.device,
batch_size=self.batch_size,
fs=self.fs,
n_fft=self.n_fft,
n_mels=self.n_mels,
win_len=self.win_len,
hop_len=self.hop_len)
self.valid = Data_Engin(method=self.method,
mono=self.mono,
address=self.valid_addr,
spectra_type=self.spectra_type,
device=self.device,
batch_size=self.batch_size,
fs=self.fs,
n_fft=self.n_fft,
n_mels=self.n_mels,
win_len=self.win_len,
hop_len=self.hop_len)
acc = dict()
target = dict()
prediction = dict()
for model in models:
torch.cuda.empty_cache()
idx = str(model['idx'])
print('idx:', idx)
test = Data_Engin(
method=model['method'],
mono=model['mono'],
address='./dataset/dcase/evaluation_setup/modify_evaluate.csv',
spectra_type=model['spectra_type'],
device=device,
batch_size=model['batch_size'],
fs=model['fs'],
n_fft=model['n_fft'],
n_mels=model['n_mels'])
network = load_model(model['network_address'],
network_type=model['network'],
no_class=no_class)
acc[idx], target[idx], prediction[idx] = infer(network=network,
valid_data_engine=test)
for item in prediction.values():
print(item[:10])
# classes = ['silence', 'clapping', 'laughing', 'scream-shout', 'conversation', 'happy', 'angry']
# classes = ['airport', 'shopping_mall', 'metro_station', 'street_pedestrian', 'public_square', 'street_traffic', 'tram', 'bus', 'metro', 'park']
classes = [
'airport', 'bus', 'metro', 'metro_station', 'park', 'public_square',
'shopping_mall', 'street_pedestrian', 'street_traffic', 'tram'
]
no_class = len(classes)
test = Data_Engin(
method='post',
mono='diff',
address='./dataset/dcase/evaluation_setup/modify_evaluate.csv',
spectra_type='mel_spectrum',
device=device,
batch_size=16,
fs=48000,
n_fft=2048,
n_mels=128,
win_len=2048,
hop_len=204)
# model_a = VGG_M(no_class=no_class)
# model_b = DCASE_PAST(no_class=no_class)
# network = ENSEMBLE(model_a=model_a, model_b=model_b, no_class=no_class)
network = VGG_M2(no_class=no_class)
if torch.cuda.device_count() > 1:
print("Let's use", torch.cuda.device_count(), "GPUs!")