fastai
Looks great in 2023 - but now?!?
I guess we just have to be disappointed in the fact that this hasn't been kept updated with dependency versions. I cannot even get the first lesson with the birds to work in any notebook I've tried (Google collab, Kaggle, VS Code). It fails on the first step setting up the book (fastbook.setup_book) showing the dependency issues. Then when using the "search_images_ddg" it has a timeout error. I saw a post stating the author fixed the problem (which was an error not being handled creating an infinite loop), but I've yet to see it work.
I only have so many cycles. Seems promising, but falls short. If someone has a solution that works and I could go through the training I'd love to hear about it. Otherwise, I guess I'll move on.
Thanks for the attempt!
install ddgs instead of duckduckgo_search and change the from duckduckgo_search import DDGS to from ddgs import DDGS.
If you've been hitting the rate limit, make sure to wait a day before trying again. They enforced a query limit per day unfortunately.
FAST.AI is a good and free course, if you are looking for an entry point to learn AI knowledge, I would encourage you start with this course. Even it was created a few years back it is still very helpful even today. There are some dependencies issue like numpy version etc. as those libs are updated frequently, but you can watch the video for such part of the course and move on to next code book, most part of the course code books are working fine.
I guess we just have to be disappointed in the fact that this hasn't been kept updated with dependency versions. I cannot even get the first lesson with the birds to work in any notebook I've tried (Google collab, Kaggle, VS Code). It fails on the first step setting up the book (fastbook.setup_book) showing the dependency issues. Then when using the "search_images_ddg" it has a timeout error. I saw a post stating the author fixed the problem (which was an error not being handled creating an infinite loop), but I've yet to see it work.
I only have so many cycles. Seems promising, but falls short. If someone has a solution that works and I could go through the training I'd love to hear about it. Otherwise, I guess I'll move on.
Thanks for the attempt!
I havent founf any issues working with those notebooks? For part 1 I have covered 70% of the lessons and notebooks. I just copy the kaggle notebook and work on it without changing any setting. It works fine.
Keep this the same so you wont come across any dependency issues:(this is for kaggle)
