67 lines
2.2 KiB
Python
67 lines
2.2 KiB
Python
|
# -*- coding: utf-8 -*-
|
||
|
|
|
||
|
|
from typing import Optional
|
||
|
|
|
||
|
|
from transformers.configuration_utils import PretrainedConfig
|
||
|
|
|
||
|
|
|
||
|
|
class HGRN2Config(PretrainedConfig):
|
||
|
|
|
||
|
|
model_type = 'hgrn2'
|
||
|
|
keys_to_ignore_at_inference = ['past_key_values']
|
||
|
|
|
||
|
|
def __init__(
|
||
|
|
self,
|
||
|
|
vocab_size: int = 32000,
|
||
|
|
hidden_size: int = 2048,
|
||
|
|
num_hidden_layers: int = 24,
|
||
|
|
attn_mode: str = "chunk",
|
||
|
|
num_heads: Optional[int] = None,
|
||
|
|
expand_ratio: Optional[int] = 128,
|
||
|
|
use_short_conv: bool = False,
|
||
|
|
conv_size: int = 4,
|
||
|
|
share_conv_kernel: bool = True,
|
||
|
|
use_lower_bound: bool = True,
|
||
|
|
hidden_ratio: Optional[int] = 4,
|
||
|
|
intermediate_size: Optional[int] = None,
|
||
|
|
hidden_act: str = "swish",
|
||
|
|
max_position_embeddings: int = 2048,
|
||
|
|
elementwise_affine: Optional[bool] = True,
|
||
|
|
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,
|
||
|
|
initializer_range: float = 0.02,
|
||
|
|
fuse_cross_entropy: bool = True,
|
||
|
|
**kwargs
|
||
|
|
):
|
||
|
|
self.vocab_size = vocab_size
|
||
|
|
self.max_position_embeddings = max_position_embeddings
|
||
|
|
self.hidden_size = hidden_size
|
||
|
|
self.num_hidden_layers = num_hidden_layers
|
||
|
|
self.attn_mode = attn_mode
|
||
|
|
self.num_heads = num_heads
|
||
|
|
self.expand_ratio = expand_ratio
|
||
|
|
self.use_short_conv = use_short_conv
|
||
|
|
self.conv_size = conv_size
|
||
|
|
self.share_conv_kernel = share_conv_kernel
|
||
|
|
self.use_lower_bound = use_lower_bound
|
||
|
|
self.hidden_ratio = hidden_ratio
|
||
|
|
self.intermediate_size = intermediate_size
|
||
|
|
self.hidden_act = hidden_act
|
||
|
|
self.elementwise_affine = elementwise_affine
|
||
|
|
self.norm_eps = norm_eps
|
||
|
|
self.use_cache = use_cache
|
||
|
|
self.initializer_range = initializer_range
|
||
|
|
self.fuse_cross_entropy = fuse_cross_entropy
|
||
|
|
|
||
|
|
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,
|
||
|
|
)
|