mlx-community/MiMo-Audio-Tokenizer
1137
1{2 "max_audio_seconds": 1800,3 "stride_size": 2,4 "avg_pooler": 2,5 "d_model": 1280,6 "scale_embedding": false,7 "kernel_size": 3,8 "activation_function": "gelu",9 "encoder_layers": 32,10 "encoder_skip_layer_id": 3,11 "encoder_attention_heads": 20,12 "encoder_ffn_dim": 5120,13 "encoder_causal": false,14 "encoder_attn_window_size": [15 -1,16 -117 ],18 "decoder_layers": 32,19 "decoder_attention_heads": 20,20 "decoder_ffn_dim": 5120,21 "decoder_kernel_size": 3,22 "decoder_stride_size": 2,23 "decoder_causal": true,24 "decoder_attn_window_size": [25 -1,26 -127 ],28 "nfft": 960,29 "vocoder_dim": 256,30 "vocoder_intermediate_dim": 1024,31 "vocoder_num_layers": 16,32 "n_mels": 128,33 "sampling_rate": 24000,34 "hop_length": 240,35 "window_size": 960,36 "vocoder_padding": "same",37 "fmin": 0,38 "fmax": null,39 "num_quantizers": 20,40 "codebook_size": [41 1024,42 1024,43 128,44 128,45 128,46 128,47 128,48 128,49 128,50 128,51 128,52 128,53 128,54 128,55 128,56 128,57 128,58 128,59 128,60 12861 ],62 "threshold_ema_dead_code": 2,63 "position_embedding_type": "rope",64 "rope_theta": 10000,65 "rope_type": "default",66 "ln_type": "LayerNorm",67 "vocoder_attention_heads": 16,68 "vocoder_attn_window_size": [69 40,70 1071 ]72}73 