{ "architectures": [ "MarkSLMV2ForCausalLM" ], "bos_token_id": 1, "dtype": "float32", "eos_token_id": 2, "hidden_size": 768, "intermediate_size": 3072, "kv_lora_rank": 192, "max_position_embeddings": 4096, "model_type": "mark_slm_v2", "moe_intermediate_size": 384, "mtp_depth": 1, "mtp_loss_weight": 0.3, "n_dense_layers": 2, "n_routed_experts": 16, "n_shared_experts": 1, "num_attention_heads": 12, "num_experts_per_tok": 2, "num_hidden_layers": 16, "num_key_value_heads": 4, "pad_token_id": 0, "q_lora_rank": 384, "qk_nope_head_dim": 64, "qk_rope_head_dim": 32, "rms_norm_eps": 1e-05, "rope_theta": 500000.0, "router_bias_update_rate": 0.001, "transformers_version": "4.57.1", "v_head_dim": 64, "vocab_size": 49152, "auto_map": { "AutoConfig": "configuration_mark_slm_v2.MarkSLMV2Config", "AutoModelForCausalLM": "modeling_mark_slm_v2.MarkSLMV2ForCausalLM" }, "total_parameters": 329132000, "active_parameters_per_token": 155723744, "weight_state": "EXPERIMENTAL_NARROW_CHECKPOINT", "sha256": "3c8c59547297bece64b49a08f39eac57cda23a47e47016283c5a63f3b31f174d", "quantization": "IQ4_NL", "target_platforms": [ "Ollama", "LM Studio", "llama.cpp", "PyTorch" ] }