ValueError:如果选择“bert”模型，则应该使用预处理_mode= 'bert‘，反之亦然

Question

我有个问题。我得到了下面的错误ValueError: if 'bert' is selected model, then preprocess_mode='bert' should be used and vice versa。但我看不出有什么问题。我的密码怎么了？

%%time
#Importing
import ktrain
from ktrain import text
(x_train_bert, y_train_bert), (x_val_bert, y_val_bert), preproc = text.texts_from_array(
                                                                        x_train=train_X.tolist(), y_train=train_y.tolist(),
                                                                        x_test=test_X.tolist(), y_test=test_y.tolist(),
                                                                        class_names=df_complete['forwarder_name'].unique(),
                                                                        preprocess_mode='bert',
                                                                        lang='en',
                                                                        maxlen=65,
                                                                        max_features=35000)                          


model = text.text_classifier(name='bert', train_data=(train_X, train_y), preproc=preproc)
learner = ktrain.get_learner(model,train_data=(train_X, train_y), val_data=(test_X, test_y), batch_size=6)

完全误差

model = text.text_classifier(name='bert', train_data=(train_X, train_y), preproc=preproc)
#learner = ktrain.get_learner(model,train_data=(train_X, train_y), val_data=(test_X, test_y), batch_size=6)
model = text.text_classifier(name='bert', train_data=(train_X, train_y), preproc=preproc)
#learner = ktrain.get_learner(model,train_data=(train_X, train_y), val_data=(test_X, test_y), batch_size=6)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [111], in <cell line: 1>()
----> 1 model = text.text_classifier(name='bert', train_data=(train_X, train_y), preproc=preproc)

File ~\AppData\Roaming\Python\Python39\site-packages\ktrain\text\models.py:589, in text_classifier(name, train_data, preproc, multilabel, metrics, verbose)
    585 if preproc is not None and not preproc.get_classes():
    586     raise ValueError(
    587         "preproc.get_classes() is empty, but required for text classification"
    588     )
--> 589 return _text_model(
    590     name,
    591     train_data,
    592     preproc=preproc,
    593     multilabel=multilabel,
    594     classification=True,
    595     metrics=metrics,
    596     verbose=verbose,
    597 )

File ~\AppData\Roaming\Python\Python39\site-packages\ktrain\text\models.py:109, in _text_model(name, train_data, preproc, multilabel, classification, metrics, verbose)
    107 is_bert = U.bert_data_tuple(train_data)
    108 if (is_bert and name != BERT) or (not is_bert and name == BERT):
--> 109     raise ValueError(
    110         "if '%s' is selected model, then preprocess_mode='%s' should be used and vice versa"
    111         % (BERT, BERT)
    112     )
    113 is_huggingface = U.is_huggingface(data=train_data)
    114 if (is_huggingface and name not in HUGGINGFACE_MODELS) or (
    115     not is_huggingface and name in HUGGINGFACE_MODELS
    116 ):

ValueError: if 'bert' is selected model, then preprocess_mode='bert' should be used and vice versa

Answer 1

你的代码里有个错误。问题是，您使用的是train_X和train_y (而不是对BERT进行预处理)，而不是x_train_bert和y_train_bert (为BERT处理)。

用这个代替：

model = text.text_classifier(name='bert', train_data=(x_train_bert, y_train_bert), preproc=preproc)
learner = ktrain.get_learner(model,train_data=(x_train_bert, y_train_bert), val_data=(x_val_bert, y_val_bert), batch_size=6)