Hi, I am studying this course, but the code has expired, so could you add me to this group, my wechat is neodreamer, thanks a lot.
except… better 
3 Likes
Hi Jeremy,
I have just submitted Lesson 5 subtitle translation and Lesson 4 subtitle translation correction.
Could you have a look? Thanks! @jeremy
By the way, I have been thinking the importance of fastai ML course and Linear Algebra courses, given fastai DL courses are so much more popular. Although in v3 part1 we have better way of storing sparse matrix and how to process Rossmann dataset as reasons to go to these two courses, I wonder what are the more systematic reasons for taking these two courses seriously.
As Jeremy has mentioned that 90% of the tasks we can do with DL (without ML), does it mean ML is not that important for a DL practitioner? If we want to become serious DL practitioners or even researchers, are these two courses the important material we can’t miss? and why?
我发现在构造文本分类器之后也可以用以下代码评判模型结果
interp = ClassificationInterpretation.from_learner(learn)
interp.plot_confusion_matrix(figsize=(7,7), dpi=60) #混淆矩阵
interp.most_confused(min_val=2) #最容易被分类错误的类别对
但是却没有类似于vision模块的 ClassificationInterpretation 中的 plot_top_losses的函数。如果想查看在validation上的表现结果（具体是哪些文本会被分类错误），需要怎么做呢？
1 Like
这是好问题，且需要动手实验的问题，我没有尝试过，也暂时无力回答。
建议在论坛里搜索一下，是否已有人提问且获得回答。
感觉是否可以从vision的plot_top_lossess的源代码入手，看是否能写一个相似的函数，套用到文本中去呢？不知道你是否尝试过？
1 Like
Thanks Jeremy. We all expect the part2.
谢谢回复，我再试试看~
答主你好，谢谢你的中文笔记和视频翻译，另外，请问一下，我发现第二课部分代码在kaggle中无法运行，你也在中文笔记中提到这一点，请问之后视频有无法运行的代码吗？
我发现已经有人开发了这个feature，并且昨天刚合并了PR（Adding TextClassificationInterpretation docs），现在 class TextClassificationInterpretation可以在这份文档看见了 
1 Like
The QR code doesn’t work anymore, could you send me a new one?
能找到你要的结果，很棒啊！
不过我快速看过后，没发现这里的文档是如何解决你提出的文本数据的plot_top_losses的问题的。你能给大家详细解说一下吗？
恩，谢了，已加入
建议楼主参考 网易公开课 吴恩达 机器学习课程，课程是中英双语字幕，但并不影响观看效果。
1 Like
谢谢你的建议，我也试试看
Hi @jeremy
I have so far received two of the same suggestion to provide bilingual subtitle.
Previously I thought it may affect viewing experience, but I do remember seeing bilingual subtitles which does not affect viewing experience elsewhere, particularly because we are provided with settings to change the font size of subtitles.
So, I wonder whether you accept bilingual subtitles for fastai videos. It is basically doing two subtitles (EN and CN) together.
The following font is large in editing mode, but as you know the size is no problem.
Thanks!
刚刚认真看了下，发现正如你所说的不是我想要的功能，于是自己实现了一下，现在能得到这样的结果。
很棒啊！还可以尝试写一个small example 来展示这个功能
Absolutely! 
1 Like