| { |
| "architectures": [ |
| "MarkEngineerV2ForCausalLM" |
| ], |
| "auto_map": { |
| "AutoConfig": "configuration_mark_engineer_v2_8b.MarkEngineerV2_8BConfig", |
| "AutoModelForCausalLM": "modeling_mark_engineer_v2_8b.MarkEngineerV2ForCausalLM" |
| }, |
| "model_type": "mark_engineer_v2_8b", |
| "model_name": "Mark Engineer V2 8B", |
| "model_version": "2.0.0", |
| "total_parameters": 8005905408, |
| "active_parameters_per_token": 1626500000, |
| "vocab_size": 49152, |
| "hidden_size": 2560, |
| "intermediate_size": 7680, |
| "num_hidden_layers": 32, |
| "num_attention_heads": 32, |
| "num_key_value_heads": 8, |
| "use_mla": true, |
| "q_lora_rank": 1024, |
| "kv_lora_rank": 512, |
| "qk_nope_head_dim": 64, |
| "qk_rope_head_dim": 32, |
| "v_head_dim": 80, |
| "n_dense_layers": 2, |
| "n_routed_experts": 24, |
| "n_shared_experts": 1, |
| "num_experts_per_tok": 2, |
| "moe_intermediate_size": 1254, |
| "recursive_depth_loops": 2, |
| "mtp_depth": 1, |
| "max_position_embeddings": 8192, |
| "rms_norm_eps": 1e-05, |
| "rope_theta": 500000.0, |
| "tie_word_embeddings": true, |
| "torch_dtype": "bfloat16", |
| "quantization_formats": [ |
| "IQ4_NL", |
| "Q4_K_M", |
| "Q8_0", |
| "F16" |
| ], |
| "target_platforms": [ |
| "Ollama", |
| "LM Studio", |
| "Transformers", |
| "vLLM" |
| ], |
| "evaluation_status": "TRAINED_BUT_UNTESTED", |
| "training_convergence_loss": 1.9047 |
| } |
