# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
| Tools for language models.
"""
from copy import copy
import numpy as np
import random
[docs]def apply_bert_mask(inputs, pad_mask, tokenizer):
"""
Apply BERT mask to the token_ids.
Args:
token_ids: The list of token ids.
Returns:
masked_token_ids: The list of masked token ids.
labels: The labels for traininig BERT.
"""
vocab_size = len(tokenizer.vocab)
bert_mask = np.random.uniform(size=inputs.shape) < 0.15
bert_mask &= pad_mask
masked_inputs = inputs * ~bert_mask
random_uniform = np.random.uniform(size=inputs.shape)
token_bert_mask = random_uniform < 0.8
random_bert_mask = random_uniform > 0.9
true_bert_mask = ~token_bert_mask & ~random_bert_mask
token_bert_mask = token_bert_mask & bert_mask
random_bert_mask = random_bert_mask & bert_mask
true_bert_mask = true_bert_mask & bert_mask
masked_inputs += tokenizer.mask_token_id * token_bert_mask
masked_inputs += np.random.randint(0, vocab_size, size=(inputs.shape)) * random_bert_mask
masked_inputs += inputs * true_bert_mask
labels = np.where(bert_mask, inputs, -1)
return masked_inputs, labels