Xiaoqing Zhou

the NMT infer OOM

1

**Is your feature request related to a problem? Please describe.** 1. Inference does not support automatic batch processing according to length, which leads to the OOM of excessively long sentences...

I'm curious about the comparison of the three models on downstream tasks. Has anyone tried it?

there are an plan to release the training part?

3

The stable training contribution mentioned in the document, while its code and script are not released. Is there an open source plan ?

what is the order of modules inserted by `add_mix` in Transformer?

1

Hello, I would like to ask two questions as follows: 1. what determines the location of the module inserted by `add_mix` ? If not specified, is it inserted at the...

如果能支持定义stopwords 就更好了

1

作者开发的中文版数据增强功能很好用，看了一下类实现，如果能支持stopwords 会更好，即部分词不进行增强，例如库nlpaug ``` import nlpaug.augmenter.word as naw aug_swap = naw.RandomWordAug(action="swap", stopwords=stopwords, aug_p=0.1) ```

In this paper, the author considered efficiency, while Inference is still time-consuming. This is because each sample x with a length of n will create 8n * k templates. >...

from newDict.add(self.idxToLabel[i]) to newDict.add(self.idxToLabel[i.numpy().tolist()])