分词器的batch_encode_plus方法存在问题

Question

我在batch_encode_plus标记器的方法中遇到了一个奇怪的问题。我最近从 Transformer 版本 3.3.0 切换到 4.5.1。（我正在为 NER 创建数据束）。

我有两个句子需要编码，并且我有一个句子已经被标记化的情况，但是由于这两个句子的长度不同，所以我需要pad [PAD]较短的句子才能获得一批统一的长度。

下面是我使用 3.3.0 版本的 Transformer 所做的代码

from transformers import AutoTokenizer

pretrained_model_name = 'distilbert-base-cased'
tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name, add_prefix_space=True)

sentences = ["He is an uninvited guest.", "The host of the party didn't sent him the invite."]

# here we have the complete sentences
encodings = tokenizer.batch_encode_plus(sentences, max_length=20, padding=True)
batch_token_ids, attention_masks = encodings["input_ids"], encodings["attention_mask"]
print(batch_token_ids[0])
print(tokenizer.convert_ids_to_tokens(batch_token_ids[0]))

# And the output
# [101, 1124, 1110, 1126, 8362, 1394, 5086, 1906, 3648, 119, 102, 0, 0, 0, 0]
# ['[CLS]', 'He', 'is', 'an', 'un', '##in', '##vi', '##ted', 'guest', '.', '[SEP]', '[PAD]', '[PAD]', '[PAD]', '[PAD]']

# here we have the already tokenized sentences
encodings = tokenizer.batch_encode_plus(batch_token_ids, max_length=20, padding=True, truncation=True, is_split_into_words=True, add_special_tokens=False, return_tensors="pt")

batch_token_ids, attention_masks = encodings["input_ids"], encodings["attention_mask"]
print(batch_token_ids[0])
print(tokenizer.convert_ids_to_tokens(batch_token_ids[0])) 

# And the output 
tensor([ 101, 1124, 1110, 1126, 8362, 1394, 5086, 1906, 3648,  119,  102, 0, 0, 0, 0])
['[CLS]', 'He', 'is', 'an', 'un', '##in', '##vi', '##ted', 'guest', '.', '[SEP]', '[PAD]', [PAD]', '[PAD]', '[PAD]']

但如果我尝试模仿 Transformer 版本 4.5.1 中的相同行为，我会得到不同的输出

from transformers import AutoTokenizer
    
pretrained_model_name = 'distilbert-base-cased'
tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name, add_prefix_space=True)

sentences = ["He is an uninvited guest.", "The host of the party didn't sent him the invite."]

# here we have the complete sentences
encodings = tokenizer.batch_encode_plus(sentences, max_length=20, padding=True)
batch_token_ids, attention_masks = encodings["input_ids"], encodings["attention_mask"]
print(batch_token_ids[0])
print(tokenizer.convert_ids_to_tokens(batch_token_ids[0]))

# And the output
#[101, 1124, 1110, 1126, 8362, 1394, 5086, 1906, 3648, 119, 102, 0, 0, 0, 0]
#['[CLS]', 'He', 'is', 'an', 'un', '##in', '##vi', '##ted', 'guest', '.', '[SEP]', '[PAD]', '[PAD]', '[PAD]', '[PAD]']

# here we have the already tokenized sentences, Note we cannot pass the batch_token_ids 
# to the batch_encode_plus method in the newer version, so need to convert them to token first
tokens1 = tokenizer.tokenize(sentences[0], add_special_tokens=True)
tokens2 = tokenizer.tokenize(sentences[1], add_special_tokens=True)

encodings = tokenizer.batch_encode_plus([tokens1, tokens2], max_length=20, padding=True, truncation=True, is_split_into_words=True, add_special_tokens=False, return_tensors="pt")

batch_token_ids, attention_masks = encodings["input_ids"], encodings["attention_mask"]
print(batch_token_ids[0])
print(tokenizer.convert_ids_to_tokens(batch_token_ids[0]))

# And the output (not the desired one)
tensor([  101,  1124,  1110,  1126,  8362,   108,   108,  1107,   108,   108,
          191,  1182,   108,   108, 21359,  1181,  3648,   119,   102])
['[CLS]', 'He', 'is', 'an', 'un', '#', '#', 'in', '#', '#', 'v', '##i', '#', '#', 'te', '##d', 'guest', '.', '[SEP]']

不知道如何处理这个问题，或者我在这里做错了什么。

Answer 1

您需要一个非快速分词器来使用整数标记列表。

tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name, add_prefix_space=True, use_fast=False)

use_fast在以后的版本中，flag 已默认启用。

从 HuggingFace 文档中，

batch_encode_plus(batch_text_or_text_pairs: ...)

batch_text_or_text_pairs (List[str], List[Tuple[str, str]], List[List[str]], List[Tuple[List[str], List[str]]], 对于非快速分词器，还有 List [列表[int]]、列表[元组[列表[int]、列表[int]]])