config.py

from transformers import PretrainedConfig

class FSTConfig(PretrainedConfig):
    model_type = "fst"

    def __init__(
        self,
        
        # Core
        vocab_size: int = 50257,
        hidden_size: int = 2048,
        num_hidden_layers: int = 24,
        num_attention_heads: int = 32,
        intermediate_size: int = 8192,
        max_position_embeddings: int = 2048,
        
        # Attention
        use_causal_attention: bool = True,
        use_cache: bool = True, # Disable during training

        # Initialization and Normalization
        initializer_range: float = 0.02,

        # Tokenizer
        bos_token_id: int | None = None,
        eos_token_id: int | None = None,
        pad_token_id: int | None = None,

        **kwargs,
    ):
        super().__init__(
            bos_token_id=bos_token_id,
            eos_token_id=eos_token_id,
            pad_token_id=pad_token_id,
            **kwargs,
        )

        # Core
        self.vocab_size = vocab_size
        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.intermediate_size = intermediate_size
        self.max_position_embeddings = max_position_embeddings

        # Attention
        self.use_causal_attention = use_causal_attention
        self.use_cache = use_cache

        # Initialization and Normalization
        self.initializer_range = initializer_range