{ "model_type": "longcoder", "hidden_size": 1024, "num_hidden_layers": 24, "num_attention_heads": 16, "intermediate_size": 4096, "max_position_embeddings": 4096, "hidden_act": "gelu", "initializer_range": 0.02, "layer_norm_eps": 1e-12, "vocab_size": 50257, "output_attentions": false, "output_hidden_states": false, "use_cache": true }