A Primer for Andrej Karpathy's PyTorch Neural Network Videos. Coding with Scott Walter

Andrej Karpathy has a fantastic video series building a neural network from scratch--but there are a few key assumptions he makes about people's understanding of Python, PyTorch, and the basics of neural networks that many might not understand without a bit of help. And that is what this video is for! We discuss at a higher level why tensors are important, then get into a simple example in Google Collab that you can follow along with!

Get The Elon Musk Mission right here:

For a limited time, use the code "Knows2021" to get 20% off your entire order!

**Want to get in on WeBull's awesome Crypto and stock fun? Check out this link, and get started trading stock and Crypto!

**You can help support this channel with one click! We have an Amazon Affiliate link in several countries. If you click the link for your country, anything you buy from Amazon in the next several hours gives us a small commission, and costs you nothing. Thank you!

**What do we use to shoot our videos?

**Here are a few products we've found really fun and/or useful:

Tesla Stock: TSLA

**JOWUA:
use the Promotion Code: DRKNOW5
use the Promotion Code: DRKNOW5
use the Promotion Code: DRKNOW5

**EVANNEX
If you use my discount code, KnowsEVs, you get $10 off any order over $100!

Instagram: @drknowitallknows

Intro: 0:00
PyTorch, Google Colab, and the importance of Tensors: 2:16
Programming a VERY Basic Neural Network: 31:48

[budgetaudiophilelife-long]

I DO NOT KNOW ANYTHING, BUT I APPRECIATE THOSE THAT DO 🤗👍😎💚💚💚

[garyswift]

I started coding about the same time as you (Texas Instruments 99-4A with tape drive because floppy disks didn't exist yet). I even used punch cards on a Univax machine at Auburn once. This was my first exposure to python. What stands out for me is the advancement of coding tools over the years. It's so much easier to code now and also easier to read the code. Though some parts of coding still haven't changed from my earliest days. Very cool. Thanks guys, that was great. One more thing, John: you mentioned wishing you remembered how you wrote some code years ago. That's how I feel about a program I wrote to solve generic quadratic equations in Fortran back around 1990.

[HarshColby]

Learning to round numbers is a simpler demonstration than any I've seen. I like it. It shows the concepts with an example where the right answer is immediately obvious to the human trying to grasp this stuff.

[kstaxman]

Great service to those who want to understand but don't have the tools to do so.. thanks!

[RobinHood-lzwj]

Thank you! So many times lectured assume that all their audience has knowledge that they really don’t. Having recourse to the fundamentals is hugely helpful to me as I learn Python.

[slowercuber]

first, thank you for bringing Andrej's excellent tuts to our attention. second, it would be nice to have a primer about how to set up colab to use A.K.'s micrograd and other libs. that would be more helpful for those of us who know python but don't know the mechanics of setting up colab to follow along with A.K. (simply importing micrograd or even !pip install micrograd in colab doesn't work)

[donson]

It's good but a bit basic, to get through quickly I'd go through Stanford's CS231n python/numpy primer Collab notebook. I recommend working through the collab file first and then read the tutorial or watch this video after you're done or as needed, or you'll fall asleep.

After first 40 s i think Idea behind this video is brilliant!

[donson]

I would do x = 4.**0.5 for example and that's better because the methods in python can be overloaded to support these more complex functions.

[daxmickelson]

Using f-strings for your gnarly print statement at the end. Note, my variables are snakecase not camelcase. print(
f"input: {round(current_input.item(), 2)}\t"
f"| predict 0: {round(probs[:, 0].item(), 2)}\t"
f"| predict 1: {round(probs[:, 1].item(), 2)}\t"
f"| target output: {current_output.item()}\t"
f"| correct: {correct_answer}"
)

[chsnoopy]

Thanks for all . I have followed the complete session. And i am a complete newbie ..but kind of works i think :D
Every now and then a Picture would be helpful ( a picture say more than thousands words :) ) where we tweaking what.. than the people or at least i would understand what we ar doing. Nevertheless thanks for the great input. It is a very interesting topic.

[domenicobarillari]

Hi - I would strongly caution the average reader that the use of the term "tensor" by computer scientists is actually incorrect. It began as a casual theft of that term to merely represent multidimensional arrays, and I suppose to sound cool, but to the mathematician, physicist and materials engineer it is a special mathematical object which has more than the property of exhibiting many indices ("rows") for many dimensions. It means a quantity that, while represented often by "big grids" in computations for human consumption, has importance invariances under certain transformations. For one thing, it will generally have different "matrix values" in difference reference frames. I invite any interested reader to chase that if otherwise also curious about relativity, etc.. Not at all connected to neural nets, except in those rare cases where a theoretical neurologist has gotten her head wrapped around deep inner properties of the brain, etc. Best regards.

[vinnylamoureux]

Believe it or not, I have a program that was written in Python before 2000. It is called "KeyNote" and was written as a developer's attempt to do something useful with the programming language he was learning. He is in Poland (if he is still alive). He dropped the project as it was taking up too much of his time, and put KeyNote out as freeware. It is a complete and handy way to create your own custom databases. I still use mine as a recipe collection. Amazingly, it works on Windows 10 !!!

[slowercuber]

BTW: If you find the concept of "too late" a little disturbing or unhelpful you can say: "It's better to simulate than never".

[RussAbbott]

It would be very useful if you would post this code somewhere, . Thanks.

At the end of the video, you say you post a link to the notebook, but I don't see it.

[tommyhood-yqet]

Do you have any recommendations for videos that will prep someone with enough python knowledge to get through this tutorial?

[theodoretourneux]

Can you tell us where to get the vids?

[donson]

I take back everything, after 33 mins it gets good.

[leibaleibovich]

Good tutorial. However, the font size is too small, it is difficult to follow for visually impaired.

[franciscogana]

Thanks it was great. I hope to see the sigmoid function.