| { | |
| "architectures": [ | |
| "LiquidCodecModel" | |
| ], | |
| "decoder_window": 30, | |
| "dtype": "bfloat16", | |
| "encoder_conv_dim": 32, | |
| "encoder_dim": 64, | |
| "frame_samples": 1920, | |
| "fsq_levels": [ | |
| 8, | |
| 7, | |
| 6, | |
| 6 | |
| ], | |
| "initializer_range": 0.02, | |
| "model_type": "liquid_codec", | |
| "n_q": 12, | |
| "quantizer_dim": 48, | |
| "sample_rate": 24000, | |
| "sampling_rate": 24000, | |
| "stft_frame_stack": 6, | |
| "stft_hop_length": 320, | |
| "stft_n_fft": 1280, | |
| "stft_win_length": 1280, | |
| "transformers_version": "5.19.0.dev0", | |
| "codec_intermediate": 128, | |
| "codec_num_layers": 4, | |
| "codec_num_heads": 2, | |
| "codec_num_kv_heads": 1, | |
| "codec_head_dim": 32, | |
| "auto_map": { | |
| "AutoConfig": "configuration_liquid_codec.LiquidCodecConfig", | |
| "AutoModel": "modeling_liquid_codec.LiquidCodecModel" | |
| } | |
| } | |