[Speech Separation/ValueError] v.3.3 - "speech_separation.py", line 648, in apply np.concatenate(remaining_zeros) ValueError: need at least one array to concatenate

[ai-nikolai]

Tested versions

• Reproducible in 3.3.

System information

M2 Pro, 14.3. pip install pyannote.audio[separation]

Issue description

Speech Separation Error.

When using file voxconverse_test_wav/jxydp.wav from Voxconverse test set.

.../speech_separation.py", line 648, in apply   
np.concatenate(remaining_zeros) 
ValueError: need at least one array to concatenate

Minimal reproduction example (MRE)

pipeline = Pipeline.from_pretrained(
            checkpoint_path="pyannote/speech-separation-ami-1.0",
            use_auth_token=auth_token,
        )
pipeline("jxydp.wav") #this throws the above error. 

[ai-nikolai]

@hbredin

[karl-ford]

Same too