from transformers import PretrainedConfig


class NanoGPTConfig(PretrainedConfig):
    model_type = "nanogpt"

    def __init__(
        self,
        sequence_len: int = 1024,
        vocab_size: int = 50304,
        n_layer: int = 12,
        n_head: int = 6,
        n_kv_head: int = 6,
        n_embd: int = 768,
        **kwargs,
    ):
        self.sequence_len = sequence_len
        self.vocab_size = vocab_size
        self.n_layer = n_layer
        self.n_head = n_head
        self.n_kv_head = n_kv_head
        self.n_embd = n_embd
        super().__init__(**kwargs)