python - Transformers 库中 Pegasus 模型的单词/句子的最大输入长度

Question

在 Transformers 库中，Pegasus 模型的单词和/或句子的最大输入长度是多少？我在 Pegasus 研究论文中读到最大值是 512 个标记，但那是多少个单词和/或句子？另外，能不能增加512个 token 的最大数量？

Answer 1

In the Transformers library, what is the maximum input length of words and/or sentences of the Pegasus model? It actually depends on your pretraining. You can create a pegagsus model that supports a length of 100 tokens or 10000 tokens. For example the model google/pegasus-cnn_dailymail supports 1024 tokens, while google/pegasus-xsum supports 512:

from transformers import PegasusTokenizerFast

t = PegasusTokenizerFast.from_pretrained("google/pegasus-xsum")
t2 = PegasusTokenizerFast.from_pretrained("google/pegasus-cnn_dailymail")
print(t.max_len_single_sentence)
print(t2.max_len_single_sentence)

输出:

511
1023

由于添加到每个序列的特殊 token ，数字减少了一个。

I read in the Pegasus research paper that the max was 512 tokens, but how many words and/or sentences is that?

这取决于你的词汇量。

from transformers import PegasusTokenizerFast
t = PegasusTokenizerFast.from_pretrained("google/pegasus-xsum")
print(t.tokenize('This is a test sentence'))
print("I know {} tokens".format(len(t)))

输出:

['▁This', '▁is', '▁a', '▁test', '▁sentence']
I know 96103 tokens

一个单词可以是一个token，也可以拆分成几个token:

print(t.tokenize('neuropsychiatric conditions'))

输出:

['▁neuro', 'psych', 'i', 'atric', '▁conditions']

Also, can you increase the maximum number of 512 tokens?

是的，您可以针对不同的输入长度训练具有 Pegasus 架构的模型，但这成本很高。