configuration_helion.py

"""
Helion Model Configuration
"""

from transformers import PretrainedConfig


class HelionConfig(PretrainedConfig):
    """
    Configuration class for Helion model.
    
    Args:
        vocab_size (`int`, *optional*, defaults to 100000):
            Vocabulary size of the Helion model.
        hidden_size (`int`, *optional*, defaults to 6144):
            Dimension of the hidden representations.
        intermediate_size (`int`, *optional*, defaults to 24576):
            Dimension of the MLP representations.
        num_hidden_layers (`int`, *optional*, defaults to 48):
            Number of hidden layers in the Transformer decoder.
        num_attention_heads (`int`, *optional*, defaults to 32):
            Number of attention heads for each attention layer.
        num_key_value_heads (`int`, *optional*, defaults to 8):
            Number of key-value heads for Grouped Query Attention.
        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
            The non-linear activation function.
        max_position_embeddings (`int`, *optional*, defaults to 16384):
            Maximum sequence length that the model can handle.
        initializer_range (`float`, *optional*, defaults to 0.02):
            Standard deviation of the truncated_normal_initializer.
        rms_norm_eps (`float`, *optional*, defaults to 1e-6):
            Epsilon value for RMSNorm layers.
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether to use cache for faster decoding.
        pad_token_id (`int`, *optional*, defaults to 0):
            Padding token id.
        bos_token_id (`int`, *optional*, defaults to 1):
            Beginning of stream token id.
        eos_token_id (`int`, *optional*, defaults to 2):
            End of stream token id.
        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
            Whether to tie input and output embeddings.
        rope_theta (`float`, *optional*, defaults to 10000.0):
            The base period of the RoPE embeddings.
        rope_scaling (`Dict`, *optional*):
            Dictionary containing the scaling configuration for RoPE.
        attention_bias (`bool`, *optional*, defaults to `False`):
            Whether to use bias in attention layers.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            Dropout probability for attention weights.
    """
    
    model_type = "helion"
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        vocab_size=100000,
        hidden_size=6144,
        intermediate_size=24576,
        num_hidden_layers=48,
        num_attention_heads=32,
        num_key_value_heads=8,
        hidden_act="silu",
        max_position_embeddings=16384,
        initializer_range=0.02,
        rms_norm_eps=1e-6,
        use_cache=True,
        pad_token_id=0,
        bos_token_id=1,
        eos_token_id=2,
        tie_word_embeddings=False,
        rope_theta=10000.0,
        rope_scaling=None,
        attention_bias=False,
        attention_dropout=0.0,
        **kwargs,
    ):
        self.vocab_size = vocab_size
        self.max_position_embeddings = max_position_embeddings
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        
        # GQA parameters
        if num_key_value_heads is None:
            num_key_value_heads = num_attention_heads
        self.num_key_value_heads = num_key_value_heads
        
        self.hidden_act = hidden_act
        self.initializer_range = initializer_range
        self.rms_norm_eps = rms_norm_eps
        self.use_cache = use_cache
        self.rope_theta = rope_theta
        self.rope_scaling = rope_scaling
        self.attention_bias = attention_bias
        self.attention_dropout = attention_dropout
        
        # Validate rope_scaling
        if self.rope_scaling is not None:
            if not isinstance(self.rope_scaling, dict):
                raise ValueError("`rope_scaling` must be a dictionary")
            
            required_keys = {"type", "factor"}
            if not all(key in self.rope_scaling for key in required_keys):
                raise ValueError(f"`rope_scaling` must contain keys {required_keys}")
            
            if self.rope_scaling["type"] not in ["linear", "dynamic"]:
                raise ValueError("`rope_scaling.type` must be 'linear' or 'dynamic'")

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