Part2 before Part1 again?


Hey guys, I had done part 1 of the course a few months back but did not give in the required homework time, but I did get the concepts. Now I have 1.5 months free and want to do both parts seriously. I was thinking maybe I should do Part 2 first and then revisit part 1. What this be ok or does Part 2 use material from part 1 ? I was thinking the code would make more sense.

F.Y.I: I am working as a deep learning researcher for a year so I am very familiar with DL concepts in general.

1 Like

(Phani Teja Anumanchupallik) #2

Since you worked as a deep learning researcher for a year, my suggestion is to do Part 2,and then go back to Part 1 if necessary. Part 2 covers building a deep learning library from scratch. This will make you more effective researcher in my opinion. It will teach you to write better code which will increase your productivity.


(Fabrizio) #3

Hi, Part1 is going to be quite different with Fastai v2 so I suggest you not to rewatch it.

Besides, being a DL researcher is not easy nowadays: the field is evolving quite fast, and you have to read new papers every week: time is precious!! If you have a good understanding of the basic concepts like cnn, rnn, traningWithTricks then go for Part2, do what Jeremy suggests to do. Once you complete the course pick a paper you like and experiment with it. Then keep studying hard!!

1 Like


I am taking about part 1 v3 itself



Are the part 1 concepts not needed in part2?


(Fabrizio) #6

yes I know. since you haven’t taken part 2 yet, you do not probably know that fastai (library) is under a major revision towards Fastai v2.0


(Phani Teja Anumanchupallik) #7

Mostly not needed.


(Joseph Catanzarite) #8

Hi @bluesky I think that jumping into Part 2 is ideal for someone who is, like yourself, already a deep learning practitioner. There’s a Part 2 v3 Study Group sponsored by TWiML; we meet every Saturday at 8:45 Pacific Time via Zoom. In between meetings we use the TWiML & AI Meetup Slack group, channel #fast_ai_dl We are starting Lesson 10 this Saturday. You are welcome to join us! Sign up for the meeting reminders (containing the Zoom link) and get a Slack invitation at:

You can use this Jupyter notebook to set up to run the course notebooks on Google Colab.

1 Like


Thanks! I will try to catch up. Is there a recorded or text version of previous meet ups?


(Joseph Catanzarite) #10

Recorded versions of the previous meetups can be found on the #fast_ai_dl channel of the TWiML Slack group. I’ve also shared annotated versions of the course notebooks 02_initialization and 03_minibatch. I’d focus on reading and running and understanding these notebooks rather than the recorded meetups.

To catch up:

  1. Set up (either locally or on the cloud) your infrastructure with the capability to run the course notebooks. To run on Google Colab, use the Jupyter notebook I shared above.
  2. Listen to Jeremy’s Lesson 8 and Lesson 9 lectures
  3. Review run, and play with course notebooks 00, 01, 02, and 03. I will review notebooks 04_callbacks.ipynb and 05_anneal.ipynb this Saturday, 8/10/2019.

Don’t worry if you can’t accomplish all of this by Saturday; I know it’s a lot. Just do as much as you can. You’ll get there. See you Saturday!

1 Like


Part 2.1 proved to me I had not really understood Part 1 so I did Part 1 5 and 6 again and then I did Part 2.2. As a novice it is difficult to remember the detail and at the same time I am still learning Python. I really like Part 2 because I like to understand how things work. I think Jeremy did state you should look back at the later sections.