CAPACITY_TOKEN_BUDGET = 200_000 CAPACITY_CONFIGS = [ {"label": "small", "emb_dim": 64, "n_heads": 2, "n_layers": 2}, {"label": "baseline", "emb_dim": 128, "n_heads": 4, "n_layers": 4}, {"label": "large", "emb_dim": 192, "n_heads": 6, "n_layers": 6}, ] __ __