{
    "model_type": "gpt2",
    "architectures": ["GPTForCausalLM"],
    "vocab_size": 50257,
    "vector_dim": 768,
    "context_length": 1024,
    "n_heads": 24,
    "n_layers": 20,
    "dropout_rate": 0.2,
    "qkv_bias": false,
    "activation_function": "geglu"
}