{ "d_model": 4096, "ssm_cfg": { "expand": 1, "ngroups": 32, "d_state": 128 }, "rms_norm_eps": 1e-05, "vocab_size": null, "d_inner": 4096, "d_xb": 1024, "intermediate_size": 14336, "hidden_act": "silu", "n_layer": 32, "attn_layers": [ 7, 15, 23, 31 ] }