Would've been better with a better documented library.

You'd better know python and enjoy reading man pages

I prefer to learn more about nlp in pytorch

The assigment and labs are not great :/

Grader output could be more useful.

The first two courses were OK and I was looking forward to doing the next one. The instruction for this is not bad, although most of it is about neural networks as they applied to NLP. NLP is not a major topic, honestly. But you do get some introduction to some concepts in neural networks.

The problem I have dealt with is that for this course is that the code does not work on Windows. The course uses a library called trax which has a dependency that won't work on Windows. Since I find it very helpful run all the code locally in my IDE and visually see the data and variables in the debugger instead of attempting to run everything in Jupyter, being able to run it locally is very important, at least to me.

I have spent a lot of time trying to get it to work. First on my laptop before realizing it won't work, then on an old Linux computer and my newer Mac Mini (does not work on my Mac like the first notebook claims). "Install WSL on your Windows". "For M1, all you have to do is build jaxlib from sourse. And maybe tensor flow." etc. etc. Because we all have nothing better to do than to dig into online forums trying to figure out which files to download and which commands to run, when the last thing you tried didn't work.

In summary, the videos aren't bad but if you use Windows and like to run the assignments locally, you may want to find another course.

The course is interesting, but the graded exercises are a mess. It looks like they've been recently updated and obviously they haven't been checked: Formatting errors, contradicting and confusing instructions, tests that fail because of bugs in the autograder tests... The forums are sometimes completely useless, there's a mix there of old and new information but the worst part are the "mentors". If you are lucky, you'll get one that knows their stuff, otherwise you will spend days pasting error messages, sending screenshots and trying to convince the mentor that there's actually an error there.

Everything was good except that this course uses Trax. This framework has yet to have any new releases since 2021, and I cannot manage to train deep learning models using Trax on my GPU, not even possible in Colab. Trax is also very buggy and it does not have a large community to help. Recommendation for learners: take the course after it is fully rebooted to TensorFlow unless you want to take other courses to get useful/working code for NLP.

The code assignments are poorly documented, and doesn't even follow its own instructions. E.g. the last assignment has instructions in plain text, referencing variables and how to use them. But in the comments when to actually implement the code, there are new variables and new comments which seems to not overlap with the previous instructions. This makes it very confusing when just trying to understand what is suppsed to happen in the code.

Spends a lot of time going over tedious implementation details rather than teaching interesting NLP topics and nuances, especially in the assignments. Introduction to Trax seems to be the only saving grace, one bonus star :)))).

For having Andrew Ng's course as suggested background for this course this is a big step (read as fall) down.

The course is good as an example of code that executes tasks correctly; that is, you can see how neural networks are defined and used in Trax. However, from a pedagogical point of view, I find it quite weak. Concepts are poorly explained and notebooks consist of little more than copying and pasting previously displayed code.

Theory wise, it's a decent course. But I am paying the money to access the labs and assignments as well. But all those practical work is in 'Trax' which is a task to learn and is totally useless because no one uses that. Kindly convert the labs and assignments to tensorflow or pytorch.

The final exam of the course has a bug: a test for the function TripletLossFn(v1, v2, margin=0.25) is wrong .... only after many hours of debugging I realized that it can be passed only if you switch v1 and v2

Useless i would strongly recommend you to don't enroll this course. None of the content is properly covered. For your kind notice Andrew Ng is not the instructor of this course.

This course is simply bad and very poor in explaination and code assignment. The jupyterlab is outdated and causes frustration (no debugging). The videos are very short for what they have to explain, as if the author is teaching experts. Overall, this course is not upto the standards of deeplearning.ai which has the bar high for their courses. I recently completed other specializations from dl.ai and they are extremely well throughout. In this course, the authors seem extremely lazy in both explaining concepts and in assignments.