|
|
|
|
|
from typing import Optional |
|
|
|
from transformers.configuration_utils import PretrainedConfig |
|
|
|
|
|
class TransformerConfig(PretrainedConfig): |
|
|
|
model_type = 'transformer' |
|
keys_to_ignore_at_inference = ['past_key_values'] |
|
|
|
def __init__( |
|
self, |
|
vocab_size: int = 32000, |
|
hidden_size: int = 2048, |
|
num_hidden_layers: int = 24, |
|
num_heads: int = 32, |
|
num_kv_heads: int = None, |
|
window_size: Optional[int] = None, |
|
rope_theta: Optional[float] = 10000., |
|
max_position_embeddings: int = 2048, |
|
hidden_ratio: Optional[int] = 4, |
|
intermediate_size: Optional[int] = None, |
|
hidden_act: str = "swish", |
|
initializer_range: float = 0.02, |
|
elementwise_affine: Optional[bool] = True, |
|
norm_first: bool = False, |
|
norm_eps: float = 1e-6, |
|
use_cache: bool = True, |
|
pad_token_id: int = None, |
|
bos_token_id: int = 1, |
|
eos_token_id: int = 2, |
|
tie_word_embeddings: bool = False, |
|
attention_bias: bool = False, |
|
fuse_norm: bool = True, |
|
fuse_cross_entropy: bool = True, |
|
**kwargs, |
|
): |
|
self.vocab_size = vocab_size |
|
self.hidden_size = hidden_size |
|
self.num_hidden_layers = num_hidden_layers |
|
self.num_heads = num_heads |
|
self.num_kv_heads = num_kv_heads |
|
self.window_size = window_size |
|
self.rope_theta = rope_theta |
|
self.max_position_embeddings = max_position_embeddings |
|
|
|
self.hidden_ratio = hidden_ratio |
|
self.intermediate_size = intermediate_size |
|
self.hidden_act = hidden_act |
|
|
|
self.initializer_range = initializer_range |
|
self.elementwise_affine = elementwise_affine |
|
self.norm_first = norm_first |
|
self.norm_eps = norm_eps |
|
self.use_cache = use_cache |
|
self.attention_bias = attention_bias |
|
self.fuse_cross_entropy = fuse_cross_entropy |
|
self.fuse_norm = fuse_norm |
|
|
|
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, |
|
) |
|
|
