本文件根据../textclf/config/preprocess.py自动生成
PreprocessConfig有以下属性：
| Attribute name | Type | Default | Description |
|---|---|---|---|
| train_file | str | "train.csv" | |
| valid_file | str | "valid.csv" | |
| test_file | str | "test.csv" | |
| datadir | str | "dataset" | |
| tokenizer | str | "char" | Possible choices: space, char, nltk, jiaba |
| nwords | int | -1 | number of words to retain, -1 表示不限制字典的大小 |
| min_word_count | int | 1 | min word count in dict |