|
|
from transformers.configuration_utils import PretrainedConfig |
|
|
from transformers.utils import logging |
|
|
|
|
|
logger = logging.get_logger(__name__) |
|
|
|
|
|
class DenseTransformerConfig(PretrainedConfig): |
|
|
model_type = "dense_transformer" |
|
|
keys_to_ignore_at_inference = ["past_key_values"] |
|
|
|
|
|
def __init__( |
|
|
self, |
|
|
vocab_size=50256, |
|
|
hidden_size=768, |
|
|
intermediate_size=3072, |
|
|
num_hidden_layers=5, |
|
|
num_attention_heads=12, |
|
|
num_key_value_heads=8, |
|
|
max_position_embeddings=1024, |
|
|
rms_norm_eps=1e-6, |
|
|
attention_dropout=0.1, |
|
|
hidden_dropout=0.1, |
|
|
use_cache=True, |
|
|
pad_token_id=0, |
|
|
bos_token_id=1, |
|
|
eos_token_id=2, |
|
|
tie_word_embeddings=False, |
|
|
torch_dtype="float32", |
|
|
pretraining_tp=1, |
|
|
output_attentions=False, |
|
|
output_hidden_states=False, |
|
|
use_return_dict=True, |
|
|
|
|
|
d_head=64, |
|
|
window_size=128, |
|
|
seq_len=512, |
|
|
attn_eps=1e-6, |
|
|
ffn_eps=1e-6, |
|
|
**kwargs, |
|
|
): |
|
|
self.vocab_size = vocab_size |
|
|
self.hidden_size = hidden_size |
|
|
self.intermediate_size = intermediate_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.max_position_embeddings = max_position_embeddings |
|
|
self.rms_norm_eps = rms_norm_eps |
|
|
self.attention_dropout = attention_dropout |
|
|
self.hidden_dropout = hidden_dropout |
|
|
self.use_cache = use_cache |
|
|
self.pretraining_tp = pretraining_tp |
|
|
self.output_attentions = output_attentions |
|
|
self.output_hidden_states = output_hidden_states |
|
|
self.use_return_dict = use_return_dict |
|
|
|
|
|
|
|
|
self.d_head = d_head |
|
|
self.window_size = window_size |
|
|
self.seq_len = seq_len |
|
|
self.attn_eps = attn_eps |
|
|
self.ffn_eps = ffn_eps |
|
|
|
|
|
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, |
|
|
torch_dtype=torch_dtype, |
|
|
**kwargs, |
|
|
) |
