I’m interested in starting part 1 in early January. I think it would be great to have a little learning cohort to help with accountability and to have people to bounce ideas off of and work through problems with.
If you’re serious and interested in starting in the first or second week in January and plan to put ~10 hours/week into completing the course, reply here!
@naman-bhalla I’m not sure what you mean…it’s all already released, no?
@damien How far are you? If you wanted to join our cohort you could wait for us to catch up…
I’m open to other ideas but what I was thinking (depending on how many people are interested) would be to have a once or twice weekly chat about the progress we’re making the problems we run into. Of course, we can also make a group chat or something where we could stay in touch outside of those meeting times.
I started lesson 1. Some content of video 2,3 I already knew. So basically, I start from lecture 4. I don’t have much of time to study so we will catch up each other soon. Be aware that some content is already dated. So it will be really annoying if you get any error when try to reproduce the result e.g. two results from 2 platforms with the same code and weights.
The new course material coming would be “Part 3” i suppose.
even if the previously posted material re- Part 1 is dated, it wont hurt to learn that. The main concepts (types of layers, Python, loss, activation, etc) of DL is going to remain the same I guess.
I am finished with Lesson 2 part 1, now starting with lesson 3.
With 10 hours a week, for someone with no background in DL before is an rather optimistic estimate. I am not just rushing through the videos but rather trying to finish the assignments with minimum requirements at least and it is taking me roughly 20-25 for that so far. With Lesson 1 it took me almost a month because i also started setting up my own System. (For the record I am a developer with no experience with Python, Linux, DL so have to learn all of these at the same time).
but the good thing is, there is no class with which one has to keep the pace, rather the material is there, one can learn at one’s own pace.
@nok : thanks for sharing the link to GCP setup. Has anyone been able to customize this set up procedure with Python 3+ which is a requirement for Pytorch (from my understanding) ? I was able to install Pytorch+Anaconda+Jupyter+Cuda support on local to use with my desktop GPU, but i’m not at ease with cloud deployment. Oops just found this install script :https://gist.github.com/motiur/2e0cd3d35bc1c42b6e5d6046b65be6f4 , just need how to find out how to deploy that on GCP
I browsed the forum a bit to see the discussions on Pytorch and the upcoming course material. Here is one thread
one correction to my above post: It seems the upcoming course will be a Pytorch version of existing Part 1 (or 2?) .
There is also Pytorch version of Part1 available by a fellow student:
So this brings to the question :
should i wait to start till next material is released ?
Start learning pytorch from the already posted Part1 notebooks with Pytorch
learn via the “old” keras notebooks and when the new material is available switch to that.
I am still sticking with option 3 to continue making a foundation of concepts. So that switching to pytorch will be easier (hopefully). Switching now to Pytorch seems a bit challenge to me when there are no video lectures available. (if anyone learning DL first time with Pytorch has experience plz share it).