{ "architectures": [ "ZenyxVantaForMaskedLM" ], "model_type": "bert", "vocab_size": 151646, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "num_key_value_heads": 4, "intermediate_size": 3072, "hidden_act": "silu", "max_position_embeddings": 2048, "initializer_range": 0.02, "layer_norm_eps": 1e-06, "rope_theta": 10000.0, "torch_dtype": "bfloat16", "transformers_version": "4.45.0" }