config.json

{
    "architectures": [
        "Lake1ForAnyToAny"
    ],
    "vocab_size": 20064,
    "hidden_size": 1024,
    "num_hidden_layers": 248,
    "num_attention_heads": 128,
    "embd_pdrop": 0.0,
    "use_moe": true,
    "num_experts": 48,
    "expert_hidden_size": 248,
    "dropout": 0.0
}