This zero seems extremely fishy. I have three questions that are WORKING. I have gone back and IMPLEMENTED THEM WITH HIGHER PRECISION. This should be MORE THAN ZERO. Something is WRONG WITH THIS LAB.
Sorry, but that’s the way the grader works: if you get any kind of exception thrown during the execution of any of the tests, then you end up with 0 points for everything. The problem is that the grader can’t complete its execution because of the exception (the assertion that “threw” in your case). Yes, they probably could have written the grader to be able to “catch” exceptions, but they didn’t and we just have to live with it.
So the solution is to figure out why that assertion is failing. Let me reply on the other thread that you started about the issues with your assignment.
Note that this is the case in every course that I’ve taken here so far, so it’s not some particular thing wrong with this assignment.
Although maybe they could have put more effort into writing the test cases such that they actually include the “catch” logic in the tests. But that would be more work of course.
Just so you know, I’m still getting zero credit after having passed the unit tests on exercises 1-4. I just tried submitting again. I messaged you again.
Correction I mean that I passed the unit tests on exercises 1-5 out of 8
And am still getting zero credit. Which is why I messaged you again, so that I can get past this point.
There is no point in submitting to the grader until you pass all the tests in the notebook. As I explained above, if you get any exceptions in the grader, then you get 0 points because the grader cannot complete its execution.
The grader will crash and give you zero points if it hits anything that causes a runtime error.
Incomplete portions of the assignment will make the grader crash.
The grader is not very robust.
Thank you. With a little assistance debugging the final steps from @paulinpaloalto, I got the certificate! I truly appreciate that!!! I also really appreciate your help, especially in earlier courses in the sequence, @TMosh!
I am also really grateful for this course in general. I spent many years trying out alternative options for learning Machine Learning and Deep Learning, and I was struggling to find something that was both at my level and in my price range. I am so grateful for this opportunity, thank you to all of you. And thank you for helping me get through this final weekend here, I know this is frustrating on Memorial Day weekend, and I’m sorry I was rude.
Congratulations! It’s great to hear that you’ve completed DLS and that it was useful to you.
What’s your next step from here? Do you have enough of what you wanted to learn to start applying the techniques to problems that you want to solve? Or do you plan to take more courses?
I’m not completely sure. I probably will take more courses eventually but I need at least a short pause in between. My plan had been to try a project of some sort over the summer to get some practice and solidify the skills I’ve learned.
I have a new Ubuntu mini pc— $200– and I think that will be sufficient to let me get started with some of the simpler models or with transfer learning if I use checkpointing and potentially some longer runs. The benefit of Ubuntu and a pc instead of a laptop is that I may be able to get it to keep running rather than sleeping and stopping. We will see.
I have previously done an independent project on evolving Newtonian gravity three body orbits by numerically solving a differential equitation in python and Jupyter so I have some practice working alone. I might prefer to try spyder next though.
I had considered starting with a Kaggle data set or LIGO open data for ease of access and some sense of familiarity or if I am doing it “right” on my first independent project. I have worked with LIGO before professionally as a grad student. However in the long run I would like to branch out to something more relevant to a job, whether that’s aerospace or biomedical or chemical. I will probably take more subject relevant courses as well, at that time. I also need to work on my health, which is still a problem.
But I’ve also been trying to keep my resume fresh and my eyes open. If that magic job comes along that is the right match to my skills and that can accommodate remote part time work where I may need to leave suddenly without warning due to seizures, and where I can’t get this documented at this time, but it may be necessary to write down some communications because I have some difficulty with hearing, especially relative to those around me in my environment, then I would definitely take the job now.
I’m also continuing to study Danish, French, and a tiny bit of German, Turkish, and maybe Spanish, but I don’t expect them to be professional skills soon. I’m still working as a physics and calculus tutor though I have the summer off. A student requested I try to qualify as a differential equations tutor so I might do that, though there’s no doubt I’ve used that skill a lot in my career, it will take some work to be prepared to teach it.
Do you have any insights on this plan? Or how to improve it? Or what courses to take next when I do?
That sounds like a good plan! I don’t think I have any additional insights. Well, maybe one thing worth considering is that if you find that you need more GPU horsepower for training on any of your projects, Colab can be used in “free” mode and will still give you access to GPU/TPUs. The only downside of the non-paying mode is that you may have to wait in a queue if the servers are busy serving paying customers. Of course you will also definitely need to use checkpoints when running on Colab, because they do suspend your job after some amount of execution with no interaction. I remember someone describing tricks you can do to fool it into thinking your job is interactive, but you still get suspended eventually (single digit hours if I remember correctly, but it’s been a couple of years since I played with that).
In terms of other courses that might be relevant, I only know the curriculum here. There probably aren’t any that cover domain topics that would be relevant to physics. Maybe the TensorFlow Professional courses are worth a look if you want to get deeper into what is possible with TF just from a techniques standpoint. I haven’t taken those courses, but would be surprised if they actually offered any physics relevant case studies. Please let us know if you find any other resources or courses that are physics relevant.
If you’re interested in going deeper on TF/Keras, the TF website has lots of good tutorials and supporting material with worked examples of different kinds of solutions. It’s worth browsing through the ToC there when you’re starting on a new type of problem.
They’ve also got a huge library of pretrained models, if any part of your problem is amenable to Transfer Learning.
