Some Tips for the TensorFlow Developer Certificate Exam
As you may know, the TensorFlow team opens up the TensorFlow Developer Certificate, which takes $100 out of our pockets. Meanwhile, I keep seeing people posting great experiences taking Tensorflow in Practice course on Coursera. Then, I’m pretty sure a larger group of people would have questions about this certificate, just like me. So, here, since I just accomplished this short journey and got the certificate, I want to post some tips and thoughts, hoping to help those who are interested in getting it.
Q1. Any tips to prepare for the exam?
Here I list three recommendations for preparation:
- take the Tensorflow in Practice course!!!
- take the Tensorflow in Practice course!!!
- take the Tensorflow in Practice course!!!
Here is the link to the course: https://www.coursera.org/specializations/tensorflow-in-practice
To take the course, make sure you finish every single material or assignment by yourself and make sure you absorb whatever you accomplish. It won’t be surprising to finish the whole course in 3 weeks if you do it every day.
Q2. Is it worth taking the course or getting the certificate?
It depends.
Under the following two situations, $100 is so waste of money:
- my day-to-day work is using Tensorflow/Keras only to create deep learning frameworks
- I don’t need to prove my skill to anybody
Under the following situations, you may consider getting it:
- I’m fresh on TensorFlow implementation, and I just want to learn and get shits done
- I’m using Tensorflow/Keras in my daily work and projects, and I come from the TensorFlow 1.x.x era, I want to have a quick refresh on new features and user-friendly APIs in TensorFlow 2.x.x. For example, there is no sess.run(), and model.fit() can do everything model.fit_generator() can do, etc.
- I want to show the certificate on my profile to prove my skill
- I’m hoping to get into the certificate network
Q3. How much does a Tensorflow Developer Certificate mean?
By taking it, you should expect yourself to be able to:
- choose appropriate Keras APIs to finish data science tasks
- create a model baseline on top of which you can customize according to your project goals.
- familiar with general deep learning framework structures for different data science problems
- have no problem reading Tensorflow documents for Python
- have a good start to comfortably answer your questions by reading TensorFlow source code and help TensorFlow/Keras community.
After all, it’s an intermediate-level certificate and it’s application-oriented. So, you won’t improve much on theoretical knowledge.
Q4. I’m about to take the exam, any tips?
Read the official Candidate Handbook: https://www.tensorflow.org/site-assets/downloads/marketing/cert/TF_Certificate_Candidate_Handbook.pdf
In the handbook, make sure you also read the following materials:
- Setting_Up_TF_Developer_Certificate_Exam
- Instructions_for_taking_the_TensorFlow_Certificate_exam
Make sure your PyCharm with the most updated version is working, your GPU is working, and all the packages version meets the exam requirements.
GPU Setup
When I had problems setting up my Cuda Toolkits and CuDNN, I found the following materials helpful, thank you, Jeff, and Kai!
2020, TensorFlow 2.0 GPU (CUDA), Keras, & Python 3.7 in Windows 10 (by Jeff Heaton)
TensorFlow 2.1 doesn’t recognize my GPU, though Cuda 10.1. (by Kai)
During the exam:
Although 5 hours seem so long that you can even take 100 breaks and go to the bathroom as many times as you want, make sure you have the model baselines for all 5 questions first. Then you will have more time to improve them. This is a practical exam, so please don’t struggle to improve 5% accuracy for more than 20 minutes.
Also, one thing to remember is to use tf.keras.callbacks.ModelCheckpoint() to save your model during training to avoid unnecessarily retraining tasks. Believe me, it could frustrate you so much during the exam, especially when you solve NLP questions.
Plugin Issue:
If you can’t find the Google Developer Certificate plugin in PyCharm, you could double-check if your PyCharm version is the latest, otherwise, the plugin won’t support it.
That’s all for the TensorFlow Developer Certificate, I hope you get something from here and am willing to answer any questions.
