New model from Mistral - Pixtral 12b language vision model architecture: "dim": 5120, "n_layers": 40, "head_dim": 128, "hidden_dim": 14336, "n_heads": 32, "n_kv_heads": 8, "rope_theta": 1000000000.0, "norm_eps": 1e-05, "vocab_size": 131072, "vision_encoder": "hidden_size":