Study Group : Delhi/NCR

abhayvarma · October 23, 2018, 3:05am

Happy Morning !!

Good session, @ShubhamStark, I am using Paperspace. What about you?

Thanks,
Abhay

ShubhamStark · October 23, 2018, 3:10am

I am trying to work it on Google colab, if it doesn’t workout I will shift to one of the clouds.

noskill · October 23, 2018, 3:29am

GCP :).

p9anand · October 23, 2018, 5:57am

Hi Everyone,

I’m Prakash Anand, have 4 years of experience in Data Science. Revisiting this course. Graduated from IIT Roorkee in Applied Mathematics. Excited to be part of this group.

Thanks.

PranY · October 23, 2018, 11:02am

Hey,

This is Pranjal from Gurgaon, 3 years into Data Science, been with fastai since the start.
Excited to work together in this group.

Thanks

PranY · October 23, 2018, 11:07am

It works just as fine, only issue is that it cleans the state every 12 hours so you have to re-install and re-download anything you had earlier.

You might want to structure your code into a stateless format, meaning everything happens from scratch every time you run it. I guess, for part 1 of the course it will be just fine. When part 2 arrives, I’ll be able to share some tricks to still work it out.

My notebook for TSG challenge is a simple example for that.

abhayvarma · October 23, 2018, 3:35pm

Hey, anyone planning any meetups ?

We can meet up on Sunday’s to discuss projects and ongoing lectures.

Thanks,
Abhay

saransh661 · October 23, 2018, 3:56pm

@PranY You can connect notebook to Gdrive and save a snapshot every n epochs so you don’t have to start from scratch every time

iamkartik · October 23, 2018, 4:13pm

Colab is having some issues with the latest release of Pytorch . The notebook was giving errors when running the fit function. Setting the num_workers=0 in ImageDataBunch.from_folder() ,fixed it but the progress was really slow . Looks like will have to use a cloud provider for unfreezing layers and training them all .

saransh661 · October 23, 2018, 4:17pm

@iamkartik I suppose you meant issues with FastAI*

PranY · October 23, 2018, 4:26pm

Yeah I do that for multi step models

zhacker · October 23, 2018, 6:00pm

Yes that’s right but can we run it takes like at most 5 minutes for an epoch on googlecolab with num_workers=0

shivamchandhok · February 5, 2019, 3:31pm

The link is not working anymore.Can you share a new link

pierreguillou · March 2, 2019, 10:01pm

Hello, I’ve just started a discussion thread on fastai study groups to gather feedbacks from the organizers and participants in order to list the best practices and also to avoid some gaps.

You will find more information in this post. Thank you if you have a few minutes to participate in the discussion

KapilRaj · March 16, 2019, 5:16pm

hey any luck getting that slack link?

shivamchandhok · March 19, 2019, 3:31pm

Nope.