sheiun comments

Results 7 comments of


                                            sheiun

作者你好，我用了单轮语料进行训练，但是仍然可以进行多轮对话，所以我想问问，这个多轮对话的机制是什么，我在您代码里面看到了history[]，还有我单轮训练语料训练出来的多轮对话前后的语句存在联系性或者相关性吗？谢谢

@zcuuu 虽然我并非作者，也没有认真研读训练的方法，但就我个人的理解，使用单轮语料训练，并不会对多轮对话直接产生关连性。而多轮机制的机制，是把过去输入的 `text` 和 `bert_response` 的 `ids` 用 `[SEP]` 串接，作为 `model` 的输入，用以预测下句回复。

python 3.7、cuda10.2 pytorch1.7可以吗？为什么我这里报了很多错误

Duplicate with #50 plz search the related issues before you ask the question.

I have new GPU driver,but.....

@albbzmkh You need to make sure `torch.cuda.device_count()` is greater than `0`

fanshow: Keep some backward compatibility before FanDrawer introduction

This commit is quite significant. Even if 'FanDrawer' is proposed, the 'Chassis Fan' and 'PSU Fan' still won't have the 'drawer_name' attribute. Would using 'N/A' be better than 'Unknown' in...

Chinese Traditional Translation Progress

Polyfills

Manage role didn't work and also not throw any error

> The HierarchyException is thrown by `addRoleToMember`, before you even call queue. Are you sure that you properly handle exceptions being thrown? When I call the method `guild.addRoleToMember` it didn't...

docs: add language support

@sytone This commit explained how to display CJK correctly without download any TTF files and also let everyone know how to use their own IME to input other language. And...