Hello trend,
This is Satya Mallick from LearnOpenCV.com.
The Independence Day Sale with a massive 25% discount ends today! Click on the button below to reserve your spot on the AI rocketship.
| INDEPENDENCE DAY SALE |
Time is ticking away!
Demystifying GPU Architectures : Part 2
In our last post we learned how GPUs accelerate AI workloads.
In today's post, we will dig deeper into NVIDIA GPU architectures - Pascal, Volta, Turing, Ampere, and Hopper.
We will also learn about cuDNN and GPU performance for Deep Learning.
| Demystifying GPU Architectures : Part 2 |
And we are sharing code so you can try it out yourself.
How to own an AI interview: A neat trick!
Imagine interviewing for an AI job.
The interviewer is in a position of power because they hold the keys to your better future.
As an interviewee, you have to upset this power imbalance by proving that you hold the keys to their better future.
Most interviewers come in with an assumption that they know more than you. And yes, they probably do know more in their narrow field of expertise.
They often ask pointed questions from their domain, and it is impossible to impress them.
To shine, you have to control the conversation and move it outside their domain to a more neutral domain where you have an edge.
You need to teach the interviewer a thing or two to gain their respect.
Let's say they ask you some question about training a neural network, and you simply answer the question, you did not teach them anything new.
They will be satisfied but not impressed.
Instead, after correctly answering the question, you move the conversation to something you have recently read about. For example, you could say -
"We can also use 8-bit floating points during training. That way, there is no need to convert the model to a lower precision like INT8 during inference. The model trained can be used as is, guaranteeing absolutely no loss in performance."
That interviewer is impressed. The balance of power is shifting in your favor.
Then you go for the kill -
"Have you ever used 8FP during training?" you ask.
They probably have not. They are now on the defensive because they have to explain why they have not explored this option.
They fumble through an answer.
You cannot let them hang in discomfort. You diffuse the tension by mentioning -
"The good news is that NVIDIA's latest Hopper Architecture supports two variants of 8 bit floating point precision. I am pretty certain training frameworks will have good support for FP8 soon."
The interviewer will nod in agreement; not because they know this fact but because they are learned something new.
It's no longer an interview; it is a conversation. And by controlling the flow of the conversation, you use the limited time to show the breadth of your knowledge.
Will this work? Absolutely, but there is a catch.
The Catch
Have you ever learned how a magic trick is performed and tried it yourself?
It's a great way to make a fool of yourself.
You know the trick, but you still cannot pull it off. You lack the skills and depth that only come through practice and fundamental knowledge.
The catch in my interview trick is that you need real knowledge and depth to turn the tables on your interviewer.
You need to have the confidence to lead a conversation. Except for psychopaths, most people build such confidence by creating a body of knowledge on top of a solid foundation of conceptual understanding.
You have to know your craft in depth.
When it comes to AI our courses provide depth and breadth that will enable you to confidently crack a job interview.
We invite you to build real expertise in AI.
🇺🇸 Independence Day Sale for AI Courses by OpenCV.org ends in just a few hours.
❤️ We are giving a massive 25% discount on all our AI courses and bundles.
🎁 Save more with bundles. The best bang for the buck is the CV Master Bundle which is the collection of all our courses offered by OpenCV.org. In case CV Master too elaborate or too expensive for you, we recommend the CV DL Starter. It gives you a solid foundation is both traditional computer vision and the novel deep learning techniques.
📕 FREE book "Deep Learning with Python", Second Edition, by François Chollet with the purchase of any bundle.
🐍 FREE 3-hour course, "Python for Beginners", with the purchase of any course or bundle.
🎓 Earn a certificate of completion from OpenCV.org.
🕛 This offer is valid until midnight July 11 (11:59 PM Pacific Time).
💯 We also provide a no-questions-asked 30-day moneyback guarantee.
🦘 Let's help you jumpstart a career in AI.
| START YOUR AI CAREER |
Cheers!
Satya
Courses / YouTube / Facebook / LinkedIn / Twitter / Instagram
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.