omnivoice-mlx-q8-g64 / audio_tokenizer /preprocessor_config.json
lightsofapollo's picture
OmniVoice MLX q8-g64: Qwen3 backbone quantized via mlx-lm; Higgs tokenizer bf16
b50cab9 verified
raw
history blame contribute delete
206 Bytes
{
"feature_extractor_type": "DacFeatureExtractor",
"feature_size": 1,
"hop_length": 960,
"padding_side": "right",
"padding_value": 0.0,
"return_attention_mask": true,
"sampling_rate": 24000
}