{ "_name_or_path": "/mnt/petrelfs/zhutong/llama-moe-models/LLaMA-MoE-v1-3_5B-4_16-new", "add_weight_norm": false, "architectures": [ "LlamaMoEForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "auto_map": { "AutoConfig": "configuration_llama_moe.LlamaMoEConfig", "AutoModel": "modeling_llama_moe_hf.LlamaMoEModel", "AutoModelForCausalLM": "modeling_llama_moe_hf.LlamaMoEForCausalLM" }, "bos_token_id": 1, "calculator_type": "UniversalCalculator", "capacity_factor": 1.25, "drop_tokens": true, "dropped_padding": "zero", "eos_token_id": 2, "gate_add_noise": true, "gate_balance_loss_weight": 0.01, "gate_network": "mlp", "gate_noise_epsilon": 0.01, "gate_type": "TopKBalancedNoisyGate", "gate_use_balance": true, "gate_use_softmax": true, "gates": "mlp", "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "model_type": "llama_moe", "multiply_gate_scores": true, "num_attention_heads": 32, "num_experts": 16, "num_hidden_layers": 32, "num_key_value_heads": 32, "num_selects": 4, "pad_token_id": 0, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "score_scale_factor": 4.0, "size_experts": [ [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ], [ 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688, 688 ] ], "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.36.2", "use_cache": true, "vocab_size": 32000 }