{
    "vocab_size": 50000,
    "context_length": 512,
    "num_layers": 8,
    "num_classes": 2,
    "attn_pdrop": 0.1,
    "d_ff": 2048,
    "d_model": 512,
    "num_heads": 8,
    "residual_pdrop": 0.1
}