| """ForgePlex-M2 model configuration for Hugging Face Transformers.""" | |
| from transformers import PretrainedConfig | |
| class ForgePlexM2Config(PretrainedConfig): | |
| model_type = "forgeplex_m2" | |
| def __init__( | |
| self, | |
| vocab_size: int = 4096, | |
| hidden_size: int = 256, | |
| num_hidden_layers: int = 11, | |
| num_attention_heads: int = 8, | |
| num_key_value_heads: int = 2, | |
| head_dim: int = 32, | |
| intermediate_size: int = 707, | |
| max_position_embeddings: int = 1024, | |
| rope_theta: float = 5000.0, | |
| rms_norm_eps: float = 1e-6, | |
| tie_word_embeddings: bool = True, | |
| use_xsa_projection: bool = False, | |
| use_attn_output_gate: bool = True, | |
| use_refresh_gate: bool = True, | |
| inject_layers: list | tuple | None = None, | |
| refresh_kernel: int = 9, | |
| bos_token_id: int = 0, | |
| eos_token_id: int = 0, | |
| pad_token_id: int = 1, | |
| **kwargs, | |
| ): | |
| if inject_layers is None: | |
| inject_layers = [5, 10] | |
| 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.num_key_value_heads = num_key_value_heads | |
| self.head_dim = head_dim | |
| self.intermediate_size = intermediate_size | |
| self.max_position_embeddings = max_position_embeddings | |
| self.rope_theta = rope_theta | |
| self.rms_norm_eps = rms_norm_eps | |
| self.use_xsa_projection = use_xsa_projection | |
| self.use_attn_output_gate = use_attn_output_gate | |
| self.use_refresh_gate = use_refresh_gate | |
| self.inject_layers = list(int(i) for i in inject_layers) | |
| self.refresh_kernel = refresh_kernel | |
| super().__init__( | |
| tie_word_embeddings=tie_word_embeddings, | |
| bos_token_id=bos_token_id, | |
| eos_token_id=eos_token_id, | |
| pad_token_id=pad_token_id, | |
| **kwargs, | |
| ) | |