Hi everyone,
I realise this is a bit late judging by the dates on a lot of the forum posts but I’ve just started the course and would love to talk to others about all this great stuff I’m learning. Is anyone interested in a new study group.? I’m on lesson 3, and I’m on the east coast of Australia.
Ralph
3 Likes
Not in response to the study group question, exactly, but I’m wondering what you are using as the source for and what platform your using for the course notebooks. It seems that every direction that I go I end up with seems to be outdated execution environments. Specifically, the imports and pip installs fail because of outdated notebooks or the libraries are the wrong versions. It’s been a very frustraiting and disappointing experience for me. I hope that someone can point me in a better, more current direction.
Hi MrVon, I understand your frustration, but unfortunately cannot offer you a straightforward solution.
I am using kaggle only at the moment and hit a few deprecated dependency errors as well. But, true to be said, I have been working in the programming world for a few years now, and this is our daily life; debugging. The pace at which new developments in AI and machine learning are evolving will keep modules outdated quickly, whatever is working today will be different in 3 months. My advice, give yourself some time to learn how to fix the error messages, using Google or chatgpt, it is a very valuable long-term skill in this realm.
1 Like
Hi Wei Ni,
Thanks so much for your response. Out of a number of posts that I’ve made both on this site and on the Discord you are the first to directly respond to me.I, too, have software development experience. In fact, I may be too old to when learn this stuff. I started to develop professionally in 1981. I retired in 2017, all of this current AI stuff was in its infancy. So, yes, I get the plight of the software developer. I actually enjoy that part of the job. I’ve been very heavilly depending on a variety of LLMs for the ML learning adventure. They are why I jumped back in to developement late last year. I come from a C/C++, Pascal, Ada, Modulo 2, assembler background. Learning python has been interesting. I hadn’t worked with an interpreter for over 30 yes, back in the BASIC days on my Apple II and TRS-80s. Blah, blah, blah…
Back on topic - I did find, but have not heard back from, a guy who did rework the Duck-Duck-Go search libray for the “Is it a Bird” notebook problem. I cannot help but think that looking under enough rocks we will eventially find the gold nuggets of either how to fix the problems or someone who has actually done so and have updated notebooks. I guess we could look at this as an opportunity to be those guys who get them working… and share them with the larger community.
Yesterday I did find that Digital Oceans Paper Space has starting point notebook that is specifically for Fast.AI. Unfortunately, the only way to get to them is with a paid subscription. I’m probably going to pay the $8.00 USD, per month just to see if their notebooks actually work. If so it will be money well spent.
How far into the first course are you? Are you also reading the book, along with watching the videos?
I’d love to chat more with you. Where are you located? I’m in the US (GMT -6)
Hey MrVon, sorry to not see this one earlier. I quickly realised as a newcomer to ML that on top of mastering the basic concepts of ML I was going to have to make up for the difference in time (2 years) the hard way by having to solve all the library inconsistencies too. On top of which, following along with the book reveals how different the videos can be with respect to kaggle/colab, to name just one. I’m using the video as source mainly, then I try and go through the relevant chapter of the book, wrestling with the differences there.
I dont come from a Python background (mine’s more .Net/NodeJS/JS/Angular, etc. web development) so the first time I got stuff working locally using Conda/Jupyter Notebooks I experienced more happiness than the birth of my first child. So when anything works for me, I save it as a project and make copious notes.
1 Like
I doubt if age has much to do with it: I’m 57 and if you find ML and AI intriguing then that’s all that matters. I have much younger web dev friends, and they aren’t interested in ML at all, so you either get it or you don’t, no age barriers.
I run all my notebooks locally (using conda as the Python environment) since as a web dev the very first step has always to see something running on my own device. If that’s not working, then you have to fix it. I would never be satisfied running a notebook in Kaggle or Colab if I couldn’t get it working in localhost. In fact, it’s a real blast to train a crappy first-run model on my laptop - that’s the fastai philosophy. I’m on lesson 3, but I’d gotten up to 7 before. I left it a while, then though I better restart.
Sorry Von, looks like I have a lot to learn from you! 40 years of experience is impressive!..I should have asked before opening my mouth haha!
Come on! You know that age can only slow you down, but not stop you! I don’t want to say that age doesn’t influence us, we may get a bit slower to absorb new things (I noticed this personally), but that doesn’t stop us from learning new stuff. It takes us a bit longer, but who cares?
I am finishing the second video and have read half the book. As Raftus said in the messages above, yeah, the code can differ a bit from the source video or book.
My background is in .NET, and some scripting in Python, which is helpful for fixing error messages in this course.
In case you are still working on the search with the bird problem, here is my notebook in kaggle: https://www.kaggle.com/code/weini0/fastai-chapter-1
I am happy to chat more and help in whatever I can. I am based on the East Coast of Oz (the reason for me to check this thread), around 5 hours behind. We can do some asynchronous messaging though.
I hope your missus doesn’t read this mate, or you will be in big trouble!
Thanks for sharing your experiences, I feel the same, I have a lot to catch up on. Let’s get this done!
2 Likes