Before I get in for whom is this certification and how I cleared, let me get straight to the exam information.
- Exam needs to written in latest stable version of PyCharm IDE in your desktop.
- You will need to download the Exam Plugin from PyCharm plugins inside the IDE.
- Once you click on the plugin, a new project will be created in new venv and you need to sign in with your google account(The account you used to purchase the exam voucher. The voucher is valid for 6 months from date of purchase).
- Once venv is ready, you need to click start exam, which is 5 HRS you can always check time on top right.
- Coming to problems, you will have 5 problems like simple NN , Image Processing etc.,.
- You need not write entire code, its like writing code in HackerRank challenges, you need to complete the function.
- Evaluation is done, by the requirements or constraints set , each problem will have 5 Marks. Once you have built the model, you will need to test the model(you will get a new window to test the model once you are on respective problem file)
- So total 5 Problems , 5 points= 25 Points. Exam guidelines did not mention any cut off, but people say it is 90% which is 22.5 0R 23.
Who can clear this exam easily?
You already solved many problems in TF 2.0 either in your Job or Kaggle.
Who cannot clear this exam?(Depending on the problems given in exam)
- You are newbie learning TF2.0 ( Learn as much as implementing custom call backs and processing complex images , lamda layers etc.)
- Don’t have theoretical knowledge in Deep Learning.(You should know how the layers work internally , especially the trainable parameter calculations and setting the input size of layers)
- Don’t know how to tame real world data and feed into deep rnn’s.
Don’t worry I will share you the resources , which can help you to rip the TF2.0.
- Deep Learning Specialization by AndrewNG, Deep Learning.AI in Coursera is everything you need to know about deep learning end to end.
- Practical implementation of DL using Tensorflow is through Tensorflow Specialization by Moorey also from Deep Learning.AI.
- For more problems you can take Jose Portilla Course on Tensorflow 2.0 in Udemy.
- Practice ! Practice ! . On Kaggle you have plenty of datasets to practice.
DO’s and Don’ts :
- The constraints set in the exam , will demand a different approach to solve the problem although the problems are most common. You can solve it easily if you understand the constraints, they are really tricky but simple!
- The model needs to be present in your project directory. It all depends on your hardware to run. But you can run code in colab and after you are ok with the results you can run same here. What I did was, used my laptop for one set of hyperparams and another set using colab to run, so that time is saved.
- All that matters is time to build the model, make sure you use batches to run. I don’t think we can build model in colab and put back model file in pycharm. I never tried it, better build the model in your machine only. We have 5 hrs , if you are comfortable using TF2.0 , you can finish writing in 1 hr maximum and rest you can reserve for training and validating the models.
I have started learning Tensorflow around 9 months back and gave the exam this month. I finished in 3hrs including model submissions. I got 24/25 in model validations, we can try with different approaches, but anyway I submitted it.
Post Exam Submission:
Once you click on submit , I really don’t know if the code which we submitted will run again in their(tensorflow) machines to validate. Anyway it took around 3–4 mins to get the message successfully submitted. You will receive a mail that you have cleared the exam!
It will take 2–3 days to receive the certificate and 2 weeks to reflect on TensorFlow network.