diff --git a/src/pyannote/audio/pipelines/clustering.py b/src/pyannote/audio/pipelines/clustering.py index f3f90802..fa86eaee 100644 --- a/src/pyannote/audio/pipelines/clustering.py +++ b/src/pyannote/audio/pipelines/clustering.py @@ -664,10 +664,8 @@ class VBxClustering(BaseClustering): else: hard_clusters = np.argmax(soft_clusters, axis=2) - # re-number clusters from 0 to num_large_clusters - _, hard_clusters = np.unique(hard_clusters, return_inverse=True) hard_clusters = hard_clusters.reshape(num_chunks, num_speakers) - + return hard_clusters, soft_clusters, centroids