configuration_skywork.py

# Copyright (c) SkyworkAI and the HuggingFace Inc. team. All rights reserved.
# This code is built upon Huggingface's transformers repository.

from transformers.configuration_utils import PretrainedConfig
from transformers.utils import logging


logger = logging.get_logger(__name__)

Skywork_PRETRAINED_CONFIG_ARCHIVE_MAP = {}


class SkyworkConfig(PretrainedConfig):

    model_type = "skywork"
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        vocab_size=32000,
        hidden_size=4096,
        intermediate_size=11008,
        num_hidden_layers=32,
        num_attention_heads=32,
        num_key_value_heads=None,
        hidden_act="silu",
        max_position_embeddings=2048,
        initializer_range=0.02,
        rms_norm_eps=1e-6,
        use_cache=True,
        pad_token_id=0,
        bos_token_id=1,
        eos_token_id=2,
        pretraining_tp=1,
        tie_word_embeddings=False,
        rope_scaling=None,
        rope_theta=10000.0,
        attention_bias=False,
        use_flash_attention=False,
        **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

        # for backward compatibility
        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.pretraining_tp = pretraining_tp
        self.use_cache = use_cache
        self.rope_scaling = rope_scaling
        self.rope_theta = rope_theta
        self.attention_bias = attention_bias
        self.use_flash_attention = use_flash_attention
        if self.use_flash_attention:
            try:
                from flash_attn.flash_attn_interface import flash_attn_varlen_func
                from einops import rearrange
            except:
                raise ValueError("`use_flash_attention` requires Flash Attention 2+ and einops.\nTry `pip install einops` and installing Flash Attention from from https://github.com/Dao-AILab/flash-attention")

        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,
        )