| { |
| "model_type": "jora_mini_hybrid", |
| "architectures": [ |
| "JoraMiniForCausalLM" |
| ], |
| "version": "v11.3-shuffle", |
| "transformers_version": "5.12.1", |
| "vocab_size": 128000, |
| "hidden_size": 1024, |
| "intermediate_size": 4096, |
| "num_hidden_layers": 28, |
| "max_position_embeddings": 2048, |
| "embed_rank": 512, |
| "tie_word_embeddings": true, |
| "num_attention_heads": 8, |
| "num_key_value_heads": 2, |
| "head_dim": 128, |
| "rope_head_dim": 128, |
| "kv_lora_rank": 256, |
| "q_lora_rank": 512, |
| "mscale": 1.0, |
| "linear_num_heads": 8, |
| "linear_num_kv_heads": 2, |
| "linear_key_head_dim": 128, |
| "linear_value_head_dim": 128, |
| "linear_conv_kernel_dim": 4, |
| "linear_chunk_size": 128, |
| "full_attention_interval": 4, |
| "layer_types": [ |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention", |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention", |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention", |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention", |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention", |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention", |
| "linear_attention", |
| "linear_attention", |
| "linear_attention", |
| "full_attention" |
| ], |
| "num_full_attention_layers": 7, |
| "num_linear_attention_layers": 21, |
| "rope_theta": 100000.0, |
| "partial_rotary_factor": 0.25, |
| "hidden_act": "silu", |
| "rms_norm_eps": 1e-06, |
| "use_mtp": true, |
| "mtp_loss_weight": 0.25, |
| "mtp_shared_logits": true, |
| "use_cache": true, |
| "attention_bias": false, |
| "dtype": "bfloat16", |
| "bos_token_id": 0, |
| "eos_token_id": 1, |
| "pad_token_id": 2 |
| } |
