Original Source Here
Deep Learning / TensorFlow / Certification
Taking the TensorFlow Developer Certification
I assume that if you’re reading this article, you are either considering taking or are set to take the TensorFlow Developer Certificate exam soon. This deep learning-modelling based certification exam by Google requires you to build neural network models solely by using the TensorFlow API.
The exam uses the PyCharm IDE and thus foundational knowledge in Python and the said IDE is really essential. If you’re new to Python, it’s fine too. Just remember to do a course on the basics of Python and then you’ll be good to go. Also before taking the exam, ensure that you are familiar with the PyCharm IDE as the TensorFlow certification exam can only be taken on PyCharm and no other python platform.
Coming to the actual deep learning part of the certification, I can assure you that extensive knowledge of machine learning or even deep learning is not a requisite. I’m only saying this because that’s how it was for me. Even though I was very interested in Data Science, I wasn’t well-versed in machine learning or deep learning. However, I was able to learn the fundamentals of deep learning including training all types of neural network models from various websites, videos and free courses. It took me about two months before I was confident enough to book the exam and take it a day later.
By the way, you don’t essentially need a local GPU for the exam. The models can be trained on a modest CPU and still you’d have plenty of time left (it is a 5 hour exam). Once you have booked your certification exam which costs $100 USD, you will be sent a PDF file which takes you through the instructions from the start till the end of the exam including setting up the Pycharm environment and the TensorFlow certification exam plug-in. Be sure to go through all the instructions in the said file before starting the exam as it is really helpful and might assist you in case you run into some issues.
Keep in mind that you can apply for a TensorFlow certification exam stipend which reduces the cost of the exam to $50 USD from the official TensorFlow site. However, this application might not always get approved and even if it does, it usually takes about four weeks or so.
To get a good idea of the structure and the syllabus of the exam, you can check out this PDF provided by TensorFlow itself here. This handy documents goes through all the content which will be assessed throughout the exam. The exam consists of five parts — a regression problem, a classification problem, an image processing problem, a Natural Language Processing (NLP) problem and a time series/forecasting problem. For each section, you need to upload a single saved model to the section’s directory and click test. The certification exam’s plugin then begin to evaluate the uploaded saved model on various unseen validation data and provides a score ranging from 1 to 5–5 being the highest score for each section. Don’t worry, you can update your model and keep uploading any number of saved models to the plugin. Just keep in mind that you need to have only one saved model in the section’s directory at a given time.
Each section is more difficult than the previous section. I managed to get a score of 5/5 for the first two sections on the first try. However, for the remaining sections, on the first try, I got scores ranging from 1 to 3. Had a moment of panic but I had a lot of time left. So I kept updating the models and also created and tested new models. After two hours or so, I was able to get 5/5 score for all the five sections and boy was I happy! I clicked submit the exam and within two minutes received a mail congratulating me and stating that I had passed the TensorFlow Developer Certification exam. It doesn’t give any other feedback though.
You can also use Jupyter Notebook and/or Google Colab which has GPU access to train different models for different sections simultaneously to save a lot of time and speed up the training process. You then just need to move the saved model into Pycharm’s directory for each section and then upload the model into the plugin from there.
The TensorFlow official website is the best place to find clear tutorials on model training for each type of neural network application. The website also provides detailed information on all the available functions, features and also transfer learning models which can also be used to obtain top results. There are also other paid and free deep learning courses and tonnes of videos online which can help you gain good understanding on the model architectures and applications.
Good Luck if you’re about to take the Certification exam! Don’t sweat it. It will always be easier than what you would expect. Just remember to be consistent and make steady progress. You’ll make it.
If you want to read more articles and get full access on Medium and support me in the process, please join Medium through my affiliate link (a part of the subscription fee is used to support me directly).
Also, if you like my content, you can subscribe to my e-mail list so that you can get notified by an e-mail as soon as I publish a story.
Thank you for reading.
Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot