Hi! If anyone from Bangladesh is taking part in this course, please give a buzz.
Happy learning! 
Hi! I am from Bangladesh. Glanced over your profile and looks like we come from the same institution(NSU). Been trying to get into deep learning and all but the data preprocessing just kills me. See you around.
Add me on facebook if that suits you
Hi,
Have any of you worked with wikipedia data dumps?
I want to contribute to ULMFiT - Bangla. I’ve found no chatter regarding it so far.
Also I need some advice on how I can train NLP models on colab/kaggle because they train so slow and kernels/sessions time out.
Cheers.
Hello!
No, I have not worked with data dumps. I literally googled ULMFiT. I mostly work on the vision domain, haven’t explored NLP avenues that much.
For models which require very large amount of compute time, you can save the weights at each iteration/instance and when the session times out, you can start again from that checkpoint. You can also look at my not-so-well-documented cheat sheet here. I should warn you, this is really messy but has most of the necessary commands.
Thank you, I’ll try it out and annoy you if I get stuck anywhere.
Yeah that idea had crossed my mind, but saving and loading pickle files every few epochs is going to be a huge pain. But nobody is going to lend me unlimited GPU for free I guess 
Being broke as I am, I’ll take anything I can get my hands on.