pub trait ModelConfigLike {
// Required methods
fn max_seq_len(&self) -> usize;
fn num_layers(&self) -> usize;
fn hidden_size(&self) -> usize;
fn num_kv_heads(&self) -> usize;
fn num_attn_heads(&self) -> usize;
fn k_head_dim(&self) -> usize;
fn v_head_dim(&self) -> usize;
}