{ "pooling_mode": "structured_selfattn", "max_length": 512, "doc_max_length": 400, "skip_instruction": true, "selfattn_attn_hidden_dim": 512, "selfattn_num_hops": 8, "selfattn_output_dropout": 0.0, "selfattn_output_norm": "layernorm", "selfattn_gamma_init": 0.1, "selfattn_gamma_learnable": true }