Sentencepiece是google开源的文本Tokenzier工具,其主要原理是利用统计算法, 在语料库中生成一个类似分词器的工具,外加可以将词token化的功能。
运行成功后,命令行应该看到下面的信息:
#测试token生成,并根据token还原句子
[INFO ] - Test Tokenize
[INFO ] - Input sentence: Hello World
[INFO ] - Tokens: [▁He, ll, o, ▁, W, or, l, d]
[INFO ] - Recovered sentence: Hello World
#测试Encode生成ids,并根据ids还原句子
[INFO ] - Test Encode & Decode
[INFO ] - Input sentence: Hello World
[INFO ] - Ids: [151, 88, 21, 4, 321, 54, 31, 17]
[INFO ] - Recovered sentence: Hello World
#测试GetToken,根据id获取token
[INFO ] - Test GetToken
[INFO ] - ids: [151, 88, 21, 4, 321, 54, 31, 17]
[INFO ] - ▁He
[INFO ] - ll
[INFO ] - o
[INFO ] - ▁
[INFO ] - W
[INFO ] - or
[INFO ] - l
[INFO ] - d
#测试GetId,根据token获取id
[INFO ] - Test GetId
[INFO ] - tokens: [▁He, ll, o, ▁, W, or, l, d]
[INFO ] - 151
[INFO ] - 88
[INFO ] - 21
[INFO ] - 4
[INFO ] - 321
[INFO ] - 54
[INFO ] - 31
[INFO ] - 17