{ "architectures": [ "MixtralForCausalLM" ], "model_type": "mixtral", "hidden_size": 768, "intermediate_size": 512, "num_hidden_layers": 12, "num_attention_heads": 12, "num_key_value_heads": 4, "head_dim": 64, "num_local_experts": 24, "num_experts_per_tok": 4, "router_aux_loss_coef": 0.01, "output_router_logits": false, "vocab_size": 32000, "max_position_embeddings": 2048, "rms_norm_eps": 1e-05, "rope_theta": 10000.0, "sliding_window": null, "tie_word_embeddings": true, "hidden_act": "silu", "torch_dtype": "float32", "bos_token_id": 0, "eos_token_id": [ 7, 1 ] }