| { |
| "architectures": [ |
| "RedAE" |
| ], |
| "audio_patch_size": 480, |
| "audio_sample_rate": 24000, |
| "bottleneck_dim": 64, |
| "dec_hidden_size": 896, |
| "dec_intermediate_size": 3584, |
| "dec_max_position_embeddings": 32768, |
| "dec_max_window_layers": 0, |
| "dec_num_attention_heads": 14, |
| "dec_num_hidden_layers": 18, |
| "dec_num_key_value_heads": 2, |
| "dec_sliding_window": 64, |
| "dec_use_sliding_window": true, |
| "dtype": "float32", |
| "enc_downsample_num_hidden_layers": 4, |
| "enc_extra_downsample_rate": 2, |
| "enc_hidden_size": 896, |
| "enc_intermediate_size": 3584, |
| "enc_max_position_embeddings": 32768, |
| "enc_max_window_layers": 0, |
| "enc_num_attention_heads": 14, |
| "enc_num_hidden_layers": 18, |
| "enc_num_key_value_heads": 2, |
| "enc_sliding_window": 64, |
| "enc_use_sliding_window": true, |
| "model_type": "redae", |
| "transformers_version": "5.6.2" |
| } |
|
|