feat: add support for WeSpeaker models on Huggingface (#1466)

pyannote/audio/pipelines/speaker_verification.py

@@ -22,13 +22,16 @@
 
 import warnings
 from functools import cached_property
+from pathlib import Path
 from typing import Text, Union
 
 import numpy as np
 import torch
 import torch.nn.functional as F
 import torchaudio
 import torchaudio.compliance.kaldi as kaldi
+from huggingface_hub import hf_hub_download
+from huggingface_hub.utils import RepositoryNotFoundError
 from torch.nn.utils.rnn import pad_sequence
 
 from pyannote.audio import Inference, Model, Pipeline
@@ -395,7 +398,7 @@ class WeSpeakerPretrainedSpeakerEmbedding(BaseInference):
 
     Usage
     -----
-    >>> get_embedding = WeSpeakerPretrainedSpeakerEmbedding("wespeaker.xxxx.onnx")
+    >>> get_embedding = WeSpeakerPretrainedSpeakerEmbedding("hbredin/wespeaker-voxceleb-resnet34-LM")
     >>> assert waveforms.ndim == 3
     >>> batch_size, num_channels, num_samples = waveforms.shape
     >>> assert num_channels == 1
@@ -410,7 +413,7 @@ class WeSpeakerPretrainedSpeakerEmbedding(BaseInference):
 
     def __init__(
         self,
-        embedding: Text = "speechbrain/spkrec-ecapa-voxceleb",
+        embedding: Text = "hbredin/wespeaker-voxceleb-resnet34-LM",
         device: torch.device = None,
     ):
         if not ONNX_IS_AVAILABLE:
@@ -420,6 +423,17 @@ def __init__(
 
         super().__init__()
 
+        if not Path(embedding).exists():
+            try:
+                embedding = hf_hub_download(
+                    repo_id=embedding,
+                    filename="speaker-embedding.onnx",
+                )
+            except RepositoryNotFoundError:
+                raise ValueError(
+                    f"Could not find '{embedding}' on huggingface.co nor on local disk."
+                )
+
         self.embedding = embedding
 
         self.to(device or torch.device("cpu"))