TensorFlow in 100 Seconds

TensorFlow is a tool for machine learning capable of building deep neural networks with high-level Python code. It provides developer-friendly APIs that help software engineers train, analyze, and deploy ML models.
#programming #deeplearning #100secondsofcode
🔗 Resources
TensorFlow Docs www.tensorflow...
Fashion MNIST Tutorial www.tensorflow...
Neural Networks Overview for Data Scientists www.ibm.com/cl...
Machine Learning in 100 Seconds • Machine Learning Expla...
🔥 Get More Content - Upgrade to PRO
Upgrade to Fireship PRO at fireship.io/pro
Use code lORhwXd2 for 25% off your first payment.
🎨 My Editor Settings
Atom One Dark
vscode-icons
Fira Code Font
🔖 Topics Covered
What is TensorFlow?
How to build a neural network with TensorFlow
What is TensorFlow used for?
Who created TensorFlow?
How neural networks work
Easy neural network tutorial
What is a mathematical Tensor?

[@haxi52]

Any machine learning topic in 100 seconds... super hard, very well executed!

[@GigaChadL337]

Nope, if you don't know how neural networks works you won't understand a shit

[@connorkapooh2002]

Whilst not really appropriate as this is more of a software channel, various Deep Learning concepts explained in 100 seconds could be pretty cool too. I hardly see any people making videos on the actual intuition behind all of it and the WHY it works and how the hell did the genius discover this. Shout out 3Blue1Brown and NNFS

[@colinb8332]

3blue1brown is an incredible channel. Great shoutout

[@NoName-lu5tg]

Who is NNFS

[@mohameddaahir4]

Welch lab's "Neural Networks Demystified" is also a good one

[@kasvith]

You might wanna check @3Blue1Brown

[@bshulke]

As others mentioned, 3Blue1Brown is a great resource for learning these concepts, as well as Stat Quest with Josh Starmer.

[@DamnBoii123]

tf is TensorFlow.

[@trayambakrai]

Tensorflow is a Tensorflow.

[@AlexMJB]

Correct

[@joshbishop]

I see what you did there *nose tap*

[@EduardoConX]

true

[@3dninja54]

I agree.

[@galgrunfeld9954]

I'd love to see a longer video about it exlaining how to use TensorFlow, explaining its and neural net's basic concepts.

[@quantumastrologer5599]

kzbin.info/aero/PLcWfeUsAys2nPgh-gYRlexc6xvscdvHqX

[@juhotuho10]

you can find that elsewhere on yt, sentdex for example

[@MrAcuriteOf1337]

Step one is to use Torch instead, fuck Tensorflow

[@cgenigma675]

@@MrAcuriteOf1337 lol why is that

[@MrAcuriteOf1337]

@@cgenigma675 Torch and Tensorflow perform similarly, however Tensorflow has had a lot of design issues over the years, and is much less Pythonic to write. Just about anyone familiar with Numpy can pick up Torch and get to work, but Tensorflow is an entirely different beast. Not to mention things like Tensorflow immediately allocating all GPU memory to itself immediately, while Torch allocates only what it needs.

[@its_elkku135]

This makes ML feel approachable and incredibly useful as opposed to something obtuse and impractical, which is the way I felt about it previously.

[@TheMomentumhd]

Check out fastai. Thats their entire philosophy. ML for people without phD, zero bs aproach! I liked their courses (free as well)

[@earthling_parth]

@@TheMomentumhd thank you for exposing me to this. I didn't even know suck Ann awesome and free resource existed!

[@Yaxqb]

Hate to break it to you, but ML is obtuse and impractical if you want to do something more than hello world with mnist. I spent a good portion of my 5yr degree on ML, I questioned my sanity each 2 months. There are so many "Yes But What If" for hyperparameters, different model architectures, normalization etc. It's easy to slip up on any part of the pipeline, which invalidates your whole model. Yeah sorry for the rant i had to get this off my chest. Maybe ML is better outside academia

[@TheMomentumhd]

@@Yaxqb its quite different... often you dont need that indepth knowledge and most of the times you are just calling some already made library or amazon/google apis. But yeah I agreed with you, been doing research on deep learning and I question my sanity everyday.. the math behind lstms is currently making me insane.

[@Yaxqb]

@@TheMomentumhd in my experience, Conv1D nets were 80% as effective as lstms for time series data, and I kinda understand 1d convolution (only kinda)🤔

[@vicradon]

I'd love to see a Fireship machine learning course.

[@flyte9844]

+1

[@Noxolo1769]

Agreed, this would be awesome!

[@mikelevels1]

This would be so cool!

[@maimee1]

TensorFlow is more for specifically Deep Learning, and DL as well as ML is more academic than coding? There's all the math and knowledge you should know before using/doing these things. Not sure if that's what Fireship teaches.... or if that would be appropriate.

[@shunnie8482]

@@maimee1 This is wrong and I dont suggest anyone to do this. First go blind, learn how to build a model without knowing the math. Then once you feel comfortable and want to learn how things work, then go ahead and learn the math. The thing is, once you learn all the math behind backprop and such it seems easy but I spent weeks trying to understand it without knowing about matrices and such. If I tried to learn that math before I actually started creating models then I would've easily ran away. I dont think this is about "my way of learning", if you show people an extremely hard thing without telling them how they'll use it they will just have less discipline and will to pull through.

[@talhaibnemahmud]

Scratched my head for the last six months in the ML course at my University. Here Jeff has built the Neural Network in 100 seconds ❤️❤️❤️

[@Tri-Technology]

Same here. In my first semester ML it was only about preprocessing (I mean yeah it's important) and in the end we have seen an easy neural network for 2 minutes and that was it.

[@Piupiumacher]

@@Tri-Technology for many applications pre processing is the most important step, especially if your data is harder to digest than images.

[@Tri-Technology]

@@Piupiumacher Yes I know, but it's frustrating if you have to take 5 semester courses to finally code your first neural network. Of course preprocessing is really important, but I don't need the preprocessing if my master is not long enough to ever build something out of the preprocessed data. For me it's just a side topic and not my main domain.

[@ohyash]

Thankfully I only had to Scratch my head for 100 seconds. I didn't understand it before, I don't understand it after.

[@codewithjarred_]

Unbelievable. 10 minutes ago I had an idea that I wanted to use ML for, and now I see that he posted this! Awesome

[@lejamesbron5880]

Wow, look at the universe universing.

[@WomboBraker]

that's how good it is

[@YuriG03042]

he used tensorflow to predict what the viewers would want

[@nateo7045]

Well what's the idea?! Don't worry, none of us are smarter enough to steal it from you

[@jjjj-x9g]

They're inside your head

[@sonmangaking]

It’s crazy to see this I’ve been now out of college for two plus years and been using these frameworks for some time

[@TheOriginalJohnDoe]

Even though it's a good explanation, machine learning concepts are really hard, and things like 'activation functions', 'dense layers' and more of this terminologies are just noise if you have rarely used machine learning and understand what they do and why you want to use them.

[@copperbadge1]

Hey Jeff - Just wanted to let you know that most of your formats are very helpful in my daily professional life (even after 15-ish years of doing this. I feel old.) and just fun to watch. Thank you for putting all the effort into both delivering a great amount of info without making it feel overwhelming and radiating a positive attitude, adding to a programming culture that feels good to live and share with the people I work with. Still googling on parties what a docker container is tho. :D

[@abdelkaleem]

Your welcome

[@Aratolotl]

This was just awesome. I wish there was a full TF course made as clearly and concisely made as this video.

[@TechySpeaking]

I learned more in 100 seconds than I learned in an entire Deep Learning course in college. Granted, if I had actually paid attention in college (and this video was out then), I would have known that my final project could have been done in 20 lines of code, rather than having to re-invent the wheel.

[@maksimfedoryak]

But isn't a process of study infinite reinvention of wheel, in core?

[@_sevelin]

Never thought I would gain interest in ML back after loosing it in my university classes! Thanks!

[@omarjimenezromero3463]

well its not school usually where you do what you want o learn it like you like, so this is very often to gain interest of something of the school when one its not on the school dealing with the school system

[@prasukjain8488]

As a tensorflow developer I can say that , tf is the best deep learning library and is super simple and easy to use.

[@ABHAY-hu9kw]

Fyi, in Physics , Tensor is the quantity which is neither scalar or vector. Like current, is scalar in Ohm's law. But, vector in Maxwell's equations or in current density. So current is tensor quantity

[@moonedCake]

Love this one❤ Please Make More Videos on Machine Learning. Because you explain really really well❤❤❤

[@tobytheepic844]

I always find that you make these 100 second videos right after I start getting into the topic of the video

[@janakakumara3836]

OMG. Thank you. I finally get it. Now do one for General Relativity.

[@fakezpred]

Jax (also by google) and pytorch are great frameworks as well although they do not have something as high level as keras available built in to the framework.

[@Sakshamn-wz3fu]

torch >> tensorflow so much better to work with idk abt jax

[@shrysm]

@@Sakshamn-wz3fu why?

[@manhalrahman5785]

Wonder what happened to Theano

[@hiteshadari4790]

@@Sakshamn-wz3fu this is what I heard so i started with pytorch

[@MouliSankarS]

@@manhalrahman5785 Development ceased. There's a fork called aesara.

[@brandonortiz3349]

you should consider doing an 100 seconds of Scala programming language

[@setsunaes]

Me (heavy breathing): "I know TensorFlow" My Boss: "Prove it..."

[@ryotanada]

My old friend tensorflow. Math has been a fascinating subject for me, which means I'm naturally fixated to this topic of machine learning when I found it in university. Unfortunately, this course was one of my lowest in terms of final score. Probably because I messed up real hard in the finals. Probably gonna get a masters and try this again.

[@C_Corpze]

We need more machine learning concepts and thing explained in 100 seconds, this was very nice and helpful!

[@gurmindersingh7602]

short and fast explained. no waste of time. good work

[@patrickwhite9902]

Thank you for the graphic at 26 seconds - I needed that right now in my life!

[@thamidudharshitha5515]

Great Job Jeff. You clearly explained most fundamentals in 100 secs. My professor took months to teach this, yet I could not understand this properly. xD

[@itilin]

Huge thanks! Recently bought a book about tensorflow js. And this video gave me hope that I could handle the book.

[@thewholeworldisburningharry123]

good luck dude! i hope youll go throught it 😄

[@JacobCanote]

Cool. I worked a JS version of this. It's fun.

[@codeaviandev9049]

Great. You have a special ability to tell about anything in a limited time. Thank you.

[@aaaaanh]

I just did a presentation on my sequential model architecture yesterday to my supervisor, I wish this vid was released then so I could steal some sick visualizations 😅

[@nikhilgangaramani]

Love your 100 sec series keep it up

[@mrdbourke]

Outstanding explanation of TensorFlow!

[@Futschikatores]

are you high?

[@aaron4163]

I used tensorflow to build a real time image color sorter 5 years ago and this woulda been helpful. Thanks

[@TobiPlayHD]

Now we need (the arguably much better) PyTorch next.

[@youssef.elmoumen]

The only channel I'm checking from time to time for new videos

[@lepystudio8974]

Would love to see this over 100s and also combined with JS

[@devoxygen7395]

Your 100 sec explanation better than all other learning material even official documents

[@pinkreefinformatica1372]

Dude... the most impressive of this video is that you explain what is a tensor in 5 seconds... ;-) mayby you should do a "5 seconds series" ;-)) great job... KUTGW!!!

[@Adoroam]

would you ever consider making a fireship video about how you produce a fireship video? like, from taking notes and writing the script to animation and other visuals?

[@sonofdirt]

Yo he already did that

[@Adoroam]

@@sonofdirt lol my b. thanks. i'll go check it out.

[@maybeanonymous6846]

Fireship in 100 seconds

[@larry_berry]

Oh wow. Literally yesterday was trying to find info about tensorflow.

[@ibado5592]

While watching this vid ,got tensor flow add . Wow super relevant.

[@zafirhasananogh2421]

can proudly say never skipped a Fireship video till date, your videos are awesome. Keep up the good work!:)

[@giorgiobellisario]

Beyond 100 seconds of TensorFlow! 🚀

[@chhavikantsingh]

100 new things to take out of these 100 seconds ... Good work 👍😁

[@parker7441]

The math and stat behind these models are really cool.

[@michaszostak1105]

Some time ago I tried to understand neural networks so I watched a few ten minute videos and still knew nothing. Now I watched your 100 second python tutorial and now I understand it

[@vadiks20032]

i like these vids. i'd love videos that are "in 3600 seconds" but with same speed as the 100 seconds one, just described more in depth

[@custommarketnews]

Yes! Short fast paced videos are the best. Pls keep them coming.

[@jeffkevin3]

Will you make videos of PyTorch or JAX (or even Numpy) in the near future? I think they are very important as well as TensorFlow (or even more important when it comes to research areas instead of industries).

[@pouyan74]

Agreed 👍

[@DaEpic]

im in love with your content 👑

[@lucasteo5015]

recently, I found a cool AI framework called "Ivy", didn't get the time to try it out yet, but what it does is that it allows you to run any AI framework code on any other AI framework, it's still a work in progress but it will definitely be a must have tool someday

[@robertbinkowski]

Did a project on this. It is a massive oversimplification. But great for starters.

[@flamakespark]

Finally, I've been waiting it for ages! Ty, Jeff!

[@poglord._]

What about neural networks in 100 seconds

[@ihateorangecat]

absolutely LOVE jeff doing machine learning related contents 👏👏

[@felixc.programs8209]

Great video! Your content really inspires me to keep producing informative tech/programming/data science content on my new KZbin channel myself. Thank you for everything!

[@CodingWithLewis]

Waiting for it to be easy as: import machinelearning machinelearning.train() machinelearning.guess_if_its_a_pullover_shirt()

[@joshuamcdonald5850]

normally these 100s videos give a good overview of a topic to let me know if i should be aware of it; but this one is so filled with jargon that i doubt its really useful for anyone. neat video though, and im sure the draw factor of ML makes it a desirable topic to make a video on

[@marekcarlisle6937]

thank you bro for all these amazing tutorials that serve to get our toes wet asap

[@habib_khan]

Beautifully done mate as always.

[@lolplayfelix-_-2713]

i literally just picked up a tesla k80 card to get started with tensorflow and then my knight in shining armor shows up and makes a video about it, thanks so incredible much!

[@patrickwhite9902]

Just putting it here for anyone else, I too have a K80 and whilst it has a heap of VRAM it is an incredibly slow GPU. Slower than my old GTX 1050ti and not worth the trouble of finding old drivers and all the fiddling with settings to even get the K80 running.

[@amandeepdogra6882]

Please a video on pytorch and then a comparison b/w TF and pytorch.

[@sharecodecamp]

Now I'm ready to start my new tensorflow career 🔥💕

[@benmatt8940]

You know you've watched too much of this dude's videos when you're able to predict that would start this video by "TensorFlow an..."

[@toup0]

Any chance to get one of those power reviews for Scala 3? Would be superb!

[@PlGGS]

Please make a full course on tensorflow

[@ahasanhabibsajeeb1979]

It's been long I had been loving tensorflow

[@mitch7w]

Been waiting forever for this😍

[@draken5379]

Dude, you explained tensorflow and ml better than ive seen in 5 hour lectures or videos. You are a god, keep it up mate.

[@rishabhjain3405]

Bro it'd be great if you can create a course on tensorflow

[@climentea]

I'm a machine learning developer now! Thx!

[@oluwatobioyeleye5246]

i love the way you broke it down, you make maching learning accessible

[@huzaifasaqib8138]

Can you make a video about the best tech stacks for beginners?

[@y0Saxir]

js everything

[@olive_oak]

Have you thought about making a 100 seconds on Scala?

[@marwan.v1511]

we would like to see full course on TensorFlow 🔥

[@vectoralphaSec]

PLEASE more AI and ML videos.

[@briankithinji7302]

Just when I needed it

[@excessreactant9045]

What amazing timing. I just started learning this

[@nemeziz_prime]

Please do one for PyTorch as well xD

[@metatronicx]

You're the best and I'm tuned Fireship. Your contents are gold for me

[@sunritjana4573]

yay! this one is epic, waited for it for a while, lessgoooooo

[@Danny-qh4su]

My first thought while clicking this - “how is he gonna write hi mom!” ?

[@0xC4aE1e5]

I've just been learning AI with R, thank you.

[@ioneocla6577]

Didn't undertood a thing but it was really interesting

[@philosophiabme]

Can’t wait for the PyTorch episode 👀👀👀

[@LucidSonder]

This is going to be highly off topic. Tldr at the bottom. Ok so hear me out, remember when fireship made the video about recursion? In the outro of that video he stated (in the next video we'll talk about memoization) and he even demoed a thumbnail for the video. (Proof: I used to have an Imgur link showing the part of the video where the thumbnail was the link ends with a/JSUGE62 ) I personally cannot find evidence of that video, back when that video happened I had an exam coming up and I really wanted the video I was mad when it wasn't out on time and then I thought, I'll definitely find it before the finals... Cut to a while later, I passed and now everytime a video comes out I tell myself "it will be next" it's never come but I am hopeful. I know memoization now but I just feel like that video will staisfy something in me. I want the video. Tldr: fireship said we'll get a video about memoization, never made it. I'm sad about it.

[@886smw]

"Paypal use it as fraud detection..." The A.I. identifies PayPal as the actual fraud!🤣🤣🤣

[@asyncawait5335]

Awesome content as usual

[@stevemcwin]

Hey Jeff, could you do a video on the Nim programming language please?

[@AndrewDChristie]

this was useful, so much made clear in one video

[@rothbardfreedom]

If these clips were under 60s, KZbin would boost it crazily in Shorts

[@santiagotorres6929]

I'll add this to my resume.

[@official6686]

welp, ill get back to learning tensorflow once i have learned math 1,2,3 and calculus

[@leoaso6984]

0:23 I actually understood that sentence. I knew that B.Eng degree would be useful one day. 😅

[@isheanesunigelmisi8400]

I was literally just reading Tensorflow when this popped up

[@theandradelimited7809]

Thanks this was really helpful to understand Tensorflow

[@lemonbot7755]

fireship i just want you to know that 2 minutes and 40 secs is not equal to 100 seconds, its 160.

[@zenu903]

I'm studying Systems Dev.&Analysis and never grasped the concept of neural networks, but hearing that "think of each node as it's own linear regression" blew my mind. Is it all just basically a lot of small statistical predictions in parallel while assigning and comparing weights for how accurate each one gets?

[@benkhemismarwen2431]

this channel's purpose is to make people think "they understand" while not understanding