| { |
| "model_type": "dasheng_audiogen", |
| "architectures": [ |
| "DashengAudioGenModel" |
| ], |
| "auto_map": { |
| "AutoConfig": "configuration_dasheng_audiogen.DashengAudioGenConfig", |
| "AutoModel": "modeling_dasheng_audiogen.DashengAudioGenModel" |
| }, |
| "text_encoder_name": "google/flan-t5-large", |
| "tokenizer_name": "mispeech/dashengtokenizer", |
| "use_zero_instruction": false, |
| "instruction_seq_len": 14, |
| "task_instruction_dim": 1024, |
| "sample_rate": 16000, |
| "downsampling_ratio": 640, |
| "latent_dim": 1280, |
| "content_dim": 1024, |
| "frame_resolution": 0.005, |
| "duration_offset": 1.0, |
| "tokenizer_max_length": 512, |
| "dit_img_size": 1000, |
| "dit_patch_size": 1, |
| "dit_in_chans": 1280, |
| "dit_out_chans": 1280, |
| "dit_input_type": "1d", |
| "dit_embed_dim": 1536, |
| "dit_depth": 32, |
| "dit_num_heads": 24, |
| "dit_mlp_ratio": 4.0, |
| "dit_qk_norm": "layernorm", |
| "dit_norm_layer": "layernorm", |
| "dit_act_layer": "geglu", |
| "dit_context_norm": true, |
| "dit_time_fusion": "ada", |
| "dit_ada_sola_rank": 32, |
| "dit_ada_sola_alpha": 32, |
| "dit_ta_context_dim": 1024, |
| "dit_ta_context_fusion": "add", |
| "dit_ta_context_norm": true, |
| "dit_context_dim": 1024, |
| "dit_context_fusion": "cross", |
| "dit_context_pe_method": "none", |
| "dit_pe_method": "none", |
| "dit_rope_mode": "shared", |
| "adapter_num_heads": 16, |
| "adapter_dropout": 0.2, |
| "adapter_duration_grad_scale": 0.1, |
| "duration_predictor_filter_channels": 512, |
| "duration_predictor_n_layers": 5, |
| "duration_predictor_kernel_size": 3, |
| "duration_predictor_p_dropout": 0.5, |
| "special_tokens": [], |
| "train_special_tokens": false |
| } |
