You can take the exam with a PC. They consider training the models can take time, but 5 hours should be sufficient.
I purchased the exam with my MacBook because I need to upload a copy of my passport and take a photo with the webcam. However, I started the certification exam on my workstation with a GPU. In summary, GPU is not crucial, but speed doesn't hurt. 🐢 🐇
You need PyCharm for the exam. The exam plugin creates an environment for you.
For preparing the exam:
Better to start with a blank notebook and start coding from scratch, instead of just running the cells in the given in a tutorial.
Finding new challenges you may like. For example, if you have a binary classification problem in a tutorial, you may consider finding another problem in the real world and work on it.
Explore documentation, change the network architecture, observe how it is changing the results. If you see a new term for the first time (for example, Huber Loss function was new for me), do some quick research and check how it is working.
During the exam:
To use time efficiently, you may consider creating checkpoints (saving weights in an h5 file and initializing the weights before a new training. Note that this may not work if you change the network architecture.)
You can also use other tools if you feel more comfortable. For example, I developed the model in Jupyter Lab, and then I moved the project into PyCharm.
I hope this helps anyone who might consider taking it.