{ "component": "speaker_encoder", "attention_channels": 128, "channels": 1152, "emb_dim": 256, "eps": 1e-06, "global_context_att": true, "hop_length_ms": 10.0, "n_fft": 512, "n_mels": 80, "sample_rate": 16000, "scale": 8, "se_bottleneck_dim": 128, "win_length_ms": 25.0 }