| { | |
| "architectures": [ | |
| "CLSPModel" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_clsp.CLSPConfig", | |
| "AutoModel": "modeling_clsp.CLSPModel" | |
| }, | |
| "causal": false, | |
| "chunk_size": "-1", | |
| "cnn_module_kernel": "31,31,15,15,15,31,31", | |
| "downsampling_factor": "1,2,4,8,4,2,1", | |
| "dtype": "float32", | |
| "encoder_dim": "1280,1280,1280,1280,1280,1280,1280", | |
| "encoder_unmasked_dim": "768,768,768,768,768,768,768", | |
| "feature_dim": 128, | |
| "feedforward_dim": "3840,3840,3840,3840,3840,3840,3840", | |
| "joint_dim": 512, | |
| "left_context_frames": "-1", | |
| "model_type": "clsp", | |
| "num_encoder_layers": "1,2,3,4,1,1,1", | |
| "num_heads": "8,8,8,8,8,8,8", | |
| "output_downsampling_factor": 2, | |
| "pos_dim": 48, | |
| "pos_head_dim": "4", | |
| "query_head_dim": "32", | |
| "text_encoder_dim": 768, | |
| "transformers_version": "4.57.3", | |
| "value_head_dim": "12" | |
| } | |