| { | |
| "model_type": "amoresignal_encoder_decoder", | |
| "architecture": "weight-shared transformer encoder + weight-shared transformer decoder", | |
| "encoder": { | |
| "vocab_size": 5000, | |
| "d_model": 64, | |
| "n_heads": 4, | |
| "n_layers": 8, | |
| "ffn_dim": 256, | |
| "max_seq_len": 520, | |
| "dropout": 0.1, | |
| "total_parameters": 408392, | |
| "pad_token_id": 0, | |
| "cls_token_id": 2, | |
| "sep_token_id": 3, | |
| "mask_token_id": 4, | |
| "unk_token_id": 1 | |
| }, | |
| "decoder": { | |
| "d_model": 64, | |
| "n_heads": 4, | |
| "n_layers": 8, | |
| "ffn_dim": 256, | |
| "n_signals": 53, | |
| "n_classes": 54, | |
| "total_parameters": 79351, | |
| "confidence_source": "softmax_probability_not_trained_head" | |
| }, | |
| "combined_total_parameters": 487743 | |
| } |