| { |
| "architectures": [ |
| "Qwen3DFlyModel" |
| ], |
| "block_size": 8, |
| "confidence_head_with_markov": false, |
| "draft_vocab_size": 120832, |
| "enable_confidence_head": false, |
| "enable_hidden_correction": true, |
| "head_dim": 128, |
| "hidden_correction_intermediate_size": null, |
| "hidden_size": 4096, |
| "intermediate_size": 13312, |
| "markov_rank": 0, |
| "mask_token_id": 120023, |
| "max_position_embeddings": 262144, |
| "model_type": "qwen3", |
| "dspark_bonus_anchor": true, |
| "num_attention_heads": 64, |
| "num_hidden_layers": 5, |
| "num_key_value_heads": 8, |
| "num_target_layers": 5, |
| "rms_norm_eps": 1e-05, |
| "rope_theta": 11158840.0, |
| "target_hidden_size": 4096, |
| "target_layer_ids": [ |
| 1, |
| 20, |
| 39, |
| 58, |
| 77 |
| ], |
| "target_num_hidden_layers": 80, |
| "tie_word_embeddings": false, |
| "transformers_version": "4.57.1", |
| "vocab_size": 120832 |
| } |
|
|