|
|
|
|
|
|
|
from transformers.configuration_utils import PretrainedConfig |
|
|
|
|
|
class TAASConfig(PretrainedConfig): |
|
model_type = "TAAS" |
|
|
|
def __init__( |
|
self, |
|
hidd_dropout=0.1, |
|
intermediate_size=3072, |
|
initialize_range=0.02, |
|
max_pos_embeddings=2048, |
|
hidd_act="gelu", |
|
attention_dropout=0.1, |
|
using_task_id=True, |
|
vocabulary_size=40000, |
|
hidd_size=768, |
|
num_hidd_layers=12, |
|
layer_norm_rate=1e-05, |
|
num_atten_heads=12, |
|
pad_token_id=0, |
|
task_vocab_size=3, |
|
classifier_drop=None, |
|
pos_embedding="absolute", |
|
use_cache=True, |
|
vocab_size=4, |
|
**kwargs |
|
): |
|
super().__init__(pad_token_id=pad_token_id, **kwargs) |
|
|
|
self.vocab_size = vocabulary_size |
|
self.max_position_embeddings = max_pos_embeddings |
|
self.type_vocab_size = vocab_size |
|
self.use_task_id = using_task_id |
|
self.layer_norm_eps = layer_norm_rate |
|
self.position_embedding_type = pos_embedding |
|
self.num_attention_heads = num_atten_heads |
|
self.hidden_size = hidd_size |
|
self.attention_probs_dropout_prob = attention_dropout |
|
self.initializer_range = initialize_range |
|
self.hidden_act = hidd_act |
|
self.intermediate_size = intermediate_size |
|
self.hidden_dropout_prob = hidd_dropout |
|
self.use_cache = use_cache |
|
self.classifier_dropout = classifier_drop |
|
self.num_hidden_layers = num_hidd_layers |
|
self.task_type_vocab_size = task_vocab_size |
|
|
|
|
|
|
|
|
