|
|
|
"""Mitre model configuration""" |
|
|
|
from transformers.configuration_utils import PretrainedConfig |
|
from transformers.utils import logging |
|
|
|
|
|
logger = logging.get_logger(__name__) |
|
|
|
class MitreConfig(PretrainedConfig): |
|
model_type = "mitre" |
|
keys_to_ignore_at_inference = ["past_key_values"] |
|
attribute_map = {"num_attention_heads": "decoder_attention_heads", "hidden_size": "d_model"} |
|
|
|
def __init__( |
|
self, |
|
vocab_size=160025, |
|
max_position_embeddings=256, |
|
decoder_layers=24, |
|
decoder_ffn_dim=4096, |
|
decoder_attention_heads=16, |
|
use_cache=True, |
|
is_encoder_decoder=False, |
|
activation_function="relu", |
|
d_model=1024, |
|
dropout=0.1, |
|
attention_dropout=0.1, |
|
activation_dropout=0.0, |
|
init_std=0.02, |
|
decoder_start_token_id=2, |
|
scale_embedding=True, |
|
pad_token_id=1, |
|
bos_token_id=0, |
|
eos_token_id=2, |
|
**kwargs, |
|
): |
|
self.vocab_size = vocab_size |
|
self.max_position_embeddings = max_position_embeddings |
|
self.d_model = d_model |
|
self.decoder_ffn_dim = decoder_ffn_dim |
|
self.decoder_layers = decoder_layers |
|
self.decoder_attention_heads = decoder_attention_heads |
|
self.dropout = dropout |
|
self.attention_dropout = attention_dropout |
|
self.activation_dropout = activation_dropout |
|
self.activation_function = activation_function |
|
self.init_std = init_std |
|
self.use_cache = use_cache |
|
self.num_hidden_layers = decoder_layers |
|
self.scale_embedding = scale_embedding |
|
self.is_decoder = True |
|
self.is_encoder_decoder = False |
|
|
|
super().__init__( |
|
pad_token_id=pad_token_id, |
|
bos_token_id=bos_token_id, |
|
eos_token_id=eos_token_id, |
|
is_encoder_decoder=is_encoder_decoder, |
|
decoder_start_token_id=decoder_start_token_id, |
|
**kwargs, |
|
) |
|
|
|
MitreConfig.register_for_auto_class("AutoConfig") |