TypeError: While connecting VectorInput::data to FrameCutter::signal:

[GeorgeMcIntire]

This code produced the following error
model = maest(arch="discogs-maest-30s-pw-73e-ts")
signal, sr = librosa.load(song_file, sr =16000)
activations, labels = model.predict_labels(signal)

---

TypeError Traceback (most recent call last)
Cell In[12], line 1
----> 1 activations, labels = model.predict_labels(signal)

File ~/MAEST/models/maest.py:881, in MAEST.predict_labels(self, x)
880 def predict_labels(self, x):
--> 881 logits = self.forward(x)[0]
882 activations = nn.functional.softmax(logits, dim=-1)
883 activations = torch.mean(activations, dim=0)

File ~/MAEST/models/maest.py:830, in MAEST.forward(self, x, transformer_block, return_self_attention)
827 self.init_melspectrogram_extractor()
829 # extract melspec from raw audio
--> 830 x = self.melspectrogram_extractor(x)
831 # normalize
832 x = (x - DISCOGS_MEAN) / (DISCOGS_STD * 2)

File ~/MAEST/models/helpers/melspectrogram_extractor.py:70, in MelSpectrogramExtractor.call(self, audio)
68 # set vector input and connect the network
69 vector_input = VectorInput(audio)
---> 70 vector_input.data >> self.frameCutter.signal
72 run(vector_input)
73 mel_bands = np.array(self.pool['mel_bands'])

File /opt/miniconda3/envs/jaage/lib/python3.8/site-packages/essentia/streaming.py:59, in _StreamConnector.rshift(left, right)
55 if not right.input_algo.hasInput(right.name):
56 raise NameError('The '%s' algorithm does not have a sink called '%s''
57 %(right.input_algo.name(), right.name))
---> 59 _essentia.connect(left.output_algo, left.name,
60 right.input_algo, right.name)
62 # update connections
63 left.output_algo.connections[left].append(right)

TypeError: While connecting VectorInput::data to FrameCutter::signal:
You cannot connect more than one Source to a Sink: FrameCutter::signal is already connected to VectorInput::data`

[palonso]

Hi @GeorgeMcIntire,
Your script ran without errors on my Centos 7.5 server using a new Conda environment created following the installation steps. These are my input audio specs:

Channels       : 2
Sample Rate    : 44100
Precision      : 16-bit
Duration       : 00:02:09.02 = 5689728 samples = 9676.41 CDDA sectors
File Size      : 5.22M
Bit Rate       : 324k
Sample Encoding: MPEG audio (layer I, II or III)

Can you provide additional details? How did you install the dependencies? What type of audio are you using?