with Google Colab

(Kasper Schnack) #42

An important followup to @adityashrm21s comment:

The link will only work for fastai-students at the university - for the mooc it will not work as you need v. 0.7

installing the 0.7 version gives the error that @wontheone1 mentions. Haven’t seen any fixes yet unfortunately :confused:


( #43

Thank you I will try!.


(Gautam Doshi) #44

All of these tips by all of you are SO helpful! I haven’t started the course yet. I’m a complete beginner and wanted to ask you a rather easy question, which may seem very stupid too. Here goes: My PC is a Core i5 - 5th gen processor with 6GB RAM and it has an in-built Intel Graphics Card (NO DEDICATED NVIDIA Graphics). However, after reading that we can access Cloud GPUs like Paperspace/Google Collab, etc. I was confused whether I’d still be able to run the notebooks using just cloud GPU’s or not (on my PC which has no dedicated graphics) ?

Please do help me out, I would really appreciate it!


(Gautam Doshi) #45

Haha never mind! I just found this wonderful post intended for beginners like me:

If anyone has any additional tips then I would love it!


(Saidur Rahman) #46

Hello everyone, I’ve written this article about getting up and running with Colab for Deep Learning and Machine Learning and I’ve gotten no errors so far.

Please let me know your feedback. Thanks

1 Like

(Anmol ) #47

Do we need to run the !wget -NS --content-disposition && bash ./ command everytime the colab server disconnects due to inactivity?


(Ronaldo da Silva Alves Batista) #48

Unfortunately yes. But now there are also Kaggle Kernels available to run the script from Clouderizer. If you open a notebook and kaggle and commit the code with the code pasted in one of the cells you can run clouderizer smoothly for 9 hours. Kaggle doesn’t disconnect in this interval.

The only downsize in Kaggle compared to Colab is that the multiprocessing doesn’t work. Pytorch always throws an error. This way the code runs somewhat 1/3 slower than Colab.

I’ve created a public template in Clouderizer with fastai 1.0.x and Pytorch 1.0 up and running.

Go in Public Templates in Clouderizer. The project name is Fastai-1.0.

I hope this helps.



Unfortunately, most of the recipes out there to get Fast AI v2 course to work with Google Colab are old and don’t work in running all the code blocks in the Jupyter notebook. After a lot of trial and error, the following simple pre-processing seems to work.

Please feel free to evaluate and comment if it doesn’t work. I will be more than happy to help.

1 Like

(Gianfrancesco Angelini) #51

Hi there,
I’ve just pushed some more snippets I’ve been using on the v2 course nbs and now on the v3.
Here is the link:
Some of them are from this wonderful community.

In that repo you can find a couple of colab links too, where they are successful applied. Please let me know if you run into any issues or typos and feel free to collaborate!


(Ronaldo da Silva Alves Batista) #52

I’ve created a public template at Clouderizer which do the whole installation setup and also downloads the github repo.


Clone the project, run the snippet code in Colab or Kaggle Kernels to start the virtual machine at Clouderizer.

Now Clouderizer charges $5 a month for 200 hours of use. You can use Colab, Kaggle Kernels or any other Kernel. But the platform gives enough advantages to worth the 5 bucks.


(Wale Opakunle) #53

This helped me a lot! Thank you.


(Johan den Hartigh) #54

Yes thanks! It helped a lot