Study Group in Toronto / GTA area


(Matthew Wong) #1

Anyone here from Toronto, Canada area interested in some study group?

//mat


(Avin Regmi) #3

Hello, I’m from Toronto. Would love to share notes and ideas. We can also potentially book a room at the public library!!!


(Matthew Wong) #4

I am at the west-end also (actually Mississauga), for sure we can hang out at the public library (e.g. the one at Bloor and Yonge etc.), a nice coffee shop, or even online.

Thursday evening works the best for me in person, online weekend evening works also.

//mat


(Xu Fei) #5

Hey do you guys have a place setup yet? I’ve done part 2 v2 this year, and I want to take it seriously and study v3 together every week. Thursday evenings work for me as well and I will ask around for a more private place with a projector if it sounds good to everyone.


(Matthew Wong) #6

I don’t think we have a place yet, anywhere near subway line will be fine for me. I started watching the fast.ai vid eariler this year too, and want to get serious also (finally willing to pay for some GPU online …)


(Xu Fei) #7

I’m contacting my friends at SAS and TWG. I used to host Meetups at TWG for study sessions and they were quite generous about it. I’m sure we’ll find a sponsor soon. So far we are just 4?


(Matthew Wong) #8

oh, say hi to Geordie at TWG for me …


(Xu Fei) #9

I don’t know Geordie at TWG (it’s a pretty big company now compared to a few years ago)…

For online GPU have you tried colab? It’s not that stable but for going through the lecture material it should be enough to get an idea, and it’s free :slight_smile:

Otherwise I’d recommend the pre-emtible instances on Google Cloud Platform. They run only 24 hrs in total but you can turn them on and off to save time. Considering the ~80% discount on price it’s a great deal (for powerful GPUs like V100 you need to order in advance and wait for approval). It took me less than a day to get access.


(Matthew Wong) #10

Yup, the god blessed colab! I tried a simple fast.ai notebook on it (cats/dogs classifier I think). I tried Paperscape also, b/c last year’s Part I recommended that. I think GCP is a good pick, and I think v1 fast.ai lib is pre-configed in GCP (AWS is coming). Anyway, I think the first lecture will talk about which cloud GPU platform is suitable for the course etc., so we can make up our mind afterward. Likely will be GCP or AWS – will see. I kind want to they the TPU thou … but maybe only works well with TF and not PyTorch … will see.


(Matthew Wong) #11

seems like I got a VM working GCP!


(Xu Fei) #12

I set up these 2 events for this week’s discussions. Hope some of you guys can make it!



(Xu Fei) #13

Is it easier to set up than Paperspace? I remember installing from scratch but I imagine it’s much better with the pre-configured fastai image!


(Matthew Wong) #14

Running fast.ai v3 in GCP is pretty painless. I haven’t try paperspace this time around, but I think I am going to stick with GCP for now, maybe local setup for some data inspection etc.


(Matthew Wong) #15

waiting list already! this is popular


(Xu Fei) #16

Yeah hope you are coming at some point! This channel is quiet for now but I’m sure there are many out there wanting to learn this stuff…


(Xu Fei) #17

I think we are UTC-5 Eastern, not UTC-6 Central…


(Mike Moloch) #18

Let’s do a series of meetups for the upcoming 2nd course if possible like the one done last October?.

I’d suggest meeting 1 hour in advance before the lecture starts so people can discuss last week’s lecture and their progress on it, then watch the streaming lecture without being distracted by conversation or trying to code along.


(Anthony Bozzo) #19

Anyone still doing this in the GTA?


(Mike Moloch) #20

Not sure tbh. @shoof did the meetups for v3 last year when the course was being offered. I was hoping we could do something like this for part 2 of deep learning when it starts mid march this year.


(Xu Fei) #21

I was hoping to find a place for the course to refresh part 1 first but got too busy to do it. If anyone could help with finding a space with a projector I’d be grateful for that. Watching the videos together though could be too difficult and lengthy as we have to discuss at a different time in the same week.