Mean-pooled, float16 embeddings of
danjacobellis/audioset_opus_24kbps
from mispeech/dasheng-0.6B.
path: source clip path (string)
label: source AudioSet label indices (list of int64)
emb: 1,280-dimensional fixed-size list of float16
Audio is decoded from the source Opus bytes, mixed to mono, and resampled to
16 kHz. The embedding is the model's documented outputdim=None output:
sigmoid applied to the mean of the final… See the full description on the dataset page:
https://huggingface.co/datasets/quinnlue/audioset-dasheng-0.6b-emb.
