PyTorch Ultimate: From Basics to Cutting-Edge (Udemy.com)

Overall the course is clear and well structured. All the questions are quickly addressed by the teacher. The only downside is some sections are becoming outdated. Some of the used libraries have been deprecated. Others have been updated and the functions are slightly different from the videos. All these created errors or compatibility issues that I had to debug and fix on my own. I would also personally like to see some content on semantic segmentation for computer vision engineers.

Pros: It covers a lot of topics in a reasonable time, it can help to get a bit better overview if you already have some background in certain deep learning topics. Cons: - barely any explanation is given - therefore, one needs lots of googling to understand even small details - Uses a very old pytroch/torchvision installation (as of 2025-March), which would be useless outside of the course

It takes a little while to get used to his accent but is well worth the time. The teacher has a total grasp of the material and explains it very well. His expectation that the students put in the effort to learn lets us make much faster progress than many other courses. Of course, the student can take as long as they want so everybody wins.

I can read through any text and get a list of layers and math symbols. What I would have liked is how, why, and when the different layers, optimizers, etc. can and should be used with real-world examples.

Well put together course which was easy to follow. The exercises were interesting, the lectures covered a broad range of topics, and the coding workshops were also useful. I enjoyed the course and feel like I learnt a lot about many machine learning aspects as a result. My only piece of constructive feedback would be to say that the course could possibly be made even better by including a few more exercises that don't just essentially involve copy and pasting/re-writing code from previous lectures with minor tweaks. Some more directed exercises, like that for the GAN section would be very interesting. Thanks very much for the course Bert and all the best!

Good broad overview of PyTorch. You can skip the first chapters if already familiar with ML/DL. I liked the succinct and to-the-point teaching style. In addition, the course also covers quite some algorithms and techniques, the semi-supervised learning chapter for example I found interesting as I had not encountered the discussed approach prior. One drawback for me was that the course seems quite focused on computer vision as opposed to NLP. NLP was also covered but there seemed to be comparatively more in-depth and cutting-edge stuff related to computer vision applications.

I still like the course, but I couldn't get most of the code on the YOLO sections to work on my machine. I think the issue is primarily directory structure. As object detection is not a high priority area for me(but I find it very interesting never the less), I'm not going to take the time to figure it out now. However, more explanation and resources re exactly hoe to structure directories etc would be helpful.

I needed a course to teach me PyTorch + Lightning because I have only worked with TensorFlow + Keras so far. I can't say anything about the introductory videos to NN/DL since I skipped them and started right away with the "Tensor" section. This was perfect for me to focus on learning PyTorch. + course is well structured + instructor provides a coding example for nearly every type of neural network (NN) which I think is fantastic! + course is pretty new (recordings from late 2022) + course is kept up to date, new lectures are added + instructor replies within a day to your questions - would have loved to see more on PyTorch Lightning but that's only due to my specific use-case Big plus for using a picture of his own very cute dog as a testing image in nearly all of the lectures! He even made her a little movie star for a video classification test. I will use the provided code and videos as a reference in my future work. Thank you Bert!

This is a very comprehensive and a solid course, and having a solid background in Python would be advantageous. It would also be helpful to have some prior knowledge of machine learning in order to better understand the material. I have completed about 60% of the course and plan to give a final review once I have finished it completely. However, for those who are currently considering taking the course, it would be better to receive additional feedback. One significant benefit of the course is that the instructor is always available to answer questions related to the lectures. Once you become accustomed to the instructor's teaching style, the material becomes easier to comprehend. Therefore, if you encounter difficulties at the beginning, I recommend pushing through and continuing to learn. Occasionally, I found myself struggling with some machine learning concepts, but I was able to improve my understanding by revisiting the material and persevering. I recommend the same approach to others.

Very good course. It is compact and to the point giving you practical "templates" on how to apply different classes of DL algorithms in PyTorch. It is somewhat limited on the "why" and the intuition of certain aspects of the problem being discussed. Thus it goes well only with another more theoretical source of information or if you already have a basic understanding of the DL concepts being demonstrated. I also like the usage of VS Code, leaving out the canonical Jupyter overhead. As a suggestion to the author: please avoid in potential future videos the usage of the magnifying glass mouse pointer - it is distracting.