Big fan of PyTorch! I use it almost every day! 🔥 …and even wrote a song about it… “for an epoch in a range… do the forward pass… calculate the loss! …optimizer zero grad, lossssssss backward!” 🎤
@@codeaperture recommend what? its a library to code AI models, not models themselves. OpenAI probably USE pytorch to train their models. they are not the same. How did you watch the documentary and still not know what pytorch is
Pytorch and OpenAI are not the same. Pytorch is a LIBRARY. OpenAI probably use Pytorch. pytorch is the TOOL not the creation. OpenAI use the tool to CREATE things....
PyTorch is easily one of the best frameworks I've ever used for any purpose. From the bottom of my heart and all the other little people that you've empowered to do great things: Thank you. ❤
Was just creating an Machine Learning project and it just came up! Literally got chills learning the story of how every big tech came around to make this story successful! I just believe if I can meet you guys and contribute to pytorch, my ML journey is successful
So strange to have AMD featured so prominently in this documentary and almost no mention of Nvidia, while in a real life I don't know anyone who uses AMD gpus for machine learning and definitely absolutely no one used AMD back in the day, there was Nvidia only. Starting from original Torch that compiled itself to CUDA C code and Tensorflow doing the same and then Pytorch built on top of cuDNN. And they manage to fully ignore that side in the video, unbelievable.
@@user-nm2wc1tt9u Yeah. It just bothers me slightly to see documentaries about events I know well with such a brazen memory holing of obvious stuff everybody knows well too. Then, 20 years later there will be holy wars on some forums about how AMD brought us AI revolution via Pytorch and Nvidia was nowhere to be seen, lol.
I came here knowing very little, after Zuckerburg mentioned open sourcing during interview with Jensen Huang, to learn something about PyTorch. I appreciate that most of the people in this doco spoke of their contributions honestly, but am grateful that you highlighted the inconsistancy.
PyTorch is a successor of the Torch library, one of whose original authors was Samy Bengio. TensorFlow is a successor of Theano, one of whose original authors was Yoshua Bengio.
I love PyTorch, but I hope TensorFlow and JAX will continue to grow as well: competition is healthy. TensorFlow was focused on production, and ease-of-use came second, while PyTorch was focused on easy-of-use and production came second (this is nicely explained in the documentary). In TensorFlow 2, the usability issues were largely fixed (Keras API + eager mode by default + big clean-up + rehauled documentation), while in PyTorch 1.0, performance and production considerations started to get more love (e.g., graph mode was added). I'm a bit sad that so many people hated TF 1.x so much that they "never looked back": TF 2 is actually quite good, it's so easy to train models across multiple GPUs or on TPUs or deploy to mobile or embedded, and there's no need to install a third-party library just to get a good training loop that works in diverse environments. Anyway, just my 2 cents!
00:04 PyTorch revolutionized the AI space 03:00 Transition of Facebook AI Research to Meta AI Research and the development of Cafe 2 for mobile AI features. 08:08 PyTorch back end separated from front end for better expressiveness 10:30 PyTorch enabled easier deep learning implementation 15:07 PyTorch and Cafe 2 were merged to create a unified technical stack for research and production. 17:33 PyTorch 1.0 enabled rapid iteration speed 22:02 PyTorch adapted for hardware acceleration and community collaboration 24:08 Hardware Community support for AI deployment 28:57 PyTorch Lightning is a community project enabling innovation and democratizing AI. 31:08 PyTorch is evolving to meet industry needs
I feel very welcome to just to watch history in making, I feel so blessed. How and where you guys think I can get started to learn pytorch. If anyone knows please help. ❤
Nice video, but I find it sad there is no mention of the OG creators of torch, in particular, it is a bit disappointing there is no interview of Ronan Collobert, who contributed so much to torch and pytorch...
The more I use PyTorch the more I realize how well thought it is, if I were to make a neural network library I don't even know what I would do differently, I would just copy pytorch, although that would be a monumental task
I legit had the same experience with Pytorch, so much more intuitive compared to TensorFlow. It felt super natural when building neural networks with PyTorch
this documentary is clear evidence of how communistic use of means of production to share innovations work best than capitalistic competition trying to destroy the adversary in the market
Please don’t bake the captions into the video. If I cannot understand them, please put the transcription into the toggled closed captions. It’s distracting when not necessary like when someone is actually speaking a different language than the rest of the video (English in this case).
Everything should be open source. Innovation slows down as soon as the guys in the suits come in with their whips and start the endless pursuit of returns stressing everyone out.
But also I think and believe, almost 90%, the future would be good and better than current one (if many companies can manufacture silicon chips and there are open source models).
Microsoft and Nvidia are at the forefront of the AI revolution. SmythOS can help you tap into this power to drive innovation in your business. #AI #SmythOS
Technology goes too fast for me to keep up. I know some Python and now I just learn PyTorch exist and it’s the next BIG thing. I have the impression that only the first people who created it and use it may have some future from it.
@@ashutoshpadhi2782 I mean it was just a joke. Numpy and pytorch are similar in certain ways like handling tensor manipulation but pytorch is optimized for machine learning with gpu functionality, gradient descent and other things built in. So by all means reinvent the wheel with numpy but as I said even if you did it would take a lot longer to compute than with pytorch, lol.
![The moment we stopped understanding AI [AlexNet]](/img/n.gif)
Couldn't have become an ML engineer today without PyTorch. Thank you Adam, Soumith, and the team who created this!
Big fan of PyTorch! I use it almost every day! 🔥 …and even wrote a song about it… “for an epoch in a range… do the forward pass… calculate the loss! …optimizer zero grad, lossssssss backward!” 🎤
Nicee lyrics 👌 😅
😅
Literally watching your 24 hour pytorch vid😂
Thanks to PyTorch you could teach us deep learning 🎉
Do u know who invented pytorch ? If ure a fan u shd know that . Soumith Chintala was on of the creators :)
We got a PyTorch documentary before GTA6
When gta6 will get released and this jokecwill expire your lifecwill lose meaning.
lmao
I used to play GTA. Then I took an arrow to the knee. 😢
it would be nice if people stop this
LMFAO that's so unrelated
PyTorch is the REAL OpenAI
Do you recommend it for Healthcare?
@@codeaperture No, because of several reasons such as hallucinations, privacy, etc.
@@codeaperture recommend what? its a library to code AI models, not models themselves. OpenAI probably USE pytorch to train their models. they are not the same. How did you watch the documentary and still not know what pytorch is
Pytorch and OpenAI are not the same. Pytorch is a LIBRARY. OpenAI probably use Pytorch. pytorch is the TOOL not the creation. OpenAI use the tool to CREATE things....
PyTorch is easily one of the best frameworks I've ever used for any purpose. From the bottom of my heart and all the other little people that you've empowered to do great things: Thank you. ❤
So glad ptrblck was mentioned! Dude is a legend!
Amen!
That guy knows more about PyTorch and GPU more than ChatGPT!!
they shouldve also interviewed him
right!
The world owes these people a debt of gratitude! Well done!!
Amazing! Thank you all for building Pytorch!
We actually have a documentary on a Python Library and more than 100k people watching it in just a month of release. This is amazing guys.
PyTorch has been an invaluable tool in my AI journey. Thanks for making deep learning accessible and enjoyable!
Feeling proud as a contributor❤️
Ty for the opportunity
PyTorch is currently the BEST tool for entering the space of AI on the market! Such a valuable tool!
Was just creating an Machine Learning project and it just came up!
Literally got chills learning the story of how every big tech came around to make this story successful!
I just believe if I can meet you guys and contribute to pytorch, my ML journey is successful
Thank you, All interview's participants
So strange to have AMD featured so prominently in this documentary and almost no mention of Nvidia, while in a real life I don't know anyone who uses AMD gpus for machine learning and definitely absolutely no one used AMD back in the day, there was Nvidia only. Starting from original Torch that compiled itself to CUDA C code and Tensorflow doing the same and then Pytorch built on top of cuDNN. And they manage to fully ignore that side in the video, unbelievable.
i guess because AMD is one of the sponsors for the video so they could not mention Nvidia anywhere
@@user-nm2wc1tt9u Yeah. It just bothers me slightly to see documentaries about events I know well with such a brazen memory holing of obvious stuff everybody knows well too. Then, 20 years later there will be holy wars on some forums about how AMD brought us AI revolution via Pytorch and Nvidia was nowhere to be seen, lol.
It makes AMD and the producers look bad to push out corporate propaganda and call it a documentary. Especially about an open source project.
I came here knowing very little, after Zuckerburg mentioned open sourcing during interview with Jensen Huang, to learn something about PyTorch. I appreciate that most of the people in this doco spoke of their contributions honestly, but am grateful that you highlighted the inconsistancy.
I've been tinkering around with PyTorch for years and I've loved how beautiful this ecosystem is.
I really hope AMD keeps hiring more devs for PyTorch development, they have an incredible opportunity.
Waited months for this from the trailer to this release.
Worth it! ❤
Thank you so much from South Africa! 🇿🇦
PyTorch is a successor of the Torch library, one of whose original authors was Samy Bengio.
TensorFlow is a successor of Theano, one of whose original authors was Yoshua Bengio.
Canadian legends! Although Yoshua is now a huge AI advancement critic.
It's sad that Canada pioneered all this stuff but gave it all away to silicon valley
@@zachem66 Geoffrey Hinton sold his company to Google at auction.
Adam Paszke is a legend!!
Massive respect to the PyTorch team 💯💯
Never looked back after switching from tensorflow to PyTorch! Well done!!
I love PyTorch, but I hope TensorFlow and JAX will continue to grow as well: competition is healthy.
TensorFlow was focused on production, and ease-of-use came second, while PyTorch was focused on easy-of-use and production came second (this is nicely explained in the documentary).
In TensorFlow 2, the usability issues were largely fixed (Keras API + eager mode by default + big clean-up + rehauled documentation), while in PyTorch 1.0, performance and production considerations started to get more love (e.g., graph mode was added). I'm a bit sad that so many people hated TF 1.x so much that they "never looked back": TF 2 is actually quite good, it's so easy to train models across multiple GPUs or on TPUs or deploy to mobile or embedded, and there's no need to install a third-party library just to get a good training loop that works in diverse environments. Anyway, just my 2 cents!
Finally a video production in respectable format
Thanks for this! This is gold. (and thanks to all PyTorch contributors from past and future!)
I dont even use Pytorch so cant fake a comment but still... this is really inspirational stuff
awesome people........thank you so much for PyTorch.
I had no idea about the starting story of ONNX! Please do more of this type of content.
Amazing! So glad this hit my feed, kudos to your entire team and thank you!
Really awesome tool. It's a daily driver at this point. Love the community as well!!!
Thank you PyTorch for making our life easy
00:04 PyTorch revolutionized the AI space
03:00 Transition of Facebook AI Research to Meta AI Research and the development of Cafe 2 for mobile AI features.
08:08 PyTorch back end separated from front end for better expressiveness
10:30 PyTorch enabled easier deep learning implementation
15:07 PyTorch and Cafe 2 were merged to create a unified technical stack for research and production.
17:33 PyTorch 1.0 enabled rapid iteration speed
22:02 PyTorch adapted for hardware acceleration and community collaboration
24:08 Hardware Community support for AI deployment
28:57 PyTorch Lightning is a community project enabling innovation and democratizing AI.
31:08 PyTorch is evolving to meet industry needs
12:28 GANs = Generative Adversarial Networks - not generative autoencoders
the abbreviation is indeed off, but what he said was correct; the early stage of generative ai are autoencoders, in fact variational ones
👏🎉 Well done and congratulations on your success! Nicely made documentary as well!
I feel very welcome to just to watch history in making, I feel so blessed. How and where you guys think I can get started to learn pytorch. If anyone knows please help. ❤
Огромное спасибо за рабочую связку.
PyTorch is amazing! good job guys
Thanks for uploading this guys 🎉❤
Great documentary, so interesting to understand the history.
One of the most impactful things ive watched on UA-cam
ptrblck is an absolute legend. I'd donate that man a few dollars if he has a donate page.
Amazing! I am learning machine learning now and I will get a job in the PyTorch team in 2 years!
???
Awesome documentary, I love hearing passionate people talking about the software I use all the time. Thank you for your amazing work!
Nice video, but I find it sad there is no mention of the OG creators of torch, in particular, it is a bit disappointing there is no interview of Ronan Collobert, who contributed so much to torch and pytorch...
Mainly concentrates on corporate
This is amazing!
Thank you so much guys! ❤🔥
The more I use PyTorch the more I realize how well thought it is, if I were to make a neural network library I don't even know what I would do differently, I would just copy pytorch, although that would be a monumental task
Love this Documentary! Thank you a lot!
Wau. Just wau. So big inspiration. So many clever people in one video. Thumbs up. My dream - one day to work in team like this. 😍
lol crazy surprised, I just see Will Falcon from the PyTorch lightning documentation vids out of nowhere 😂😂😂 thank u for teaching me how DDP works
I legit had the same experience with Pytorch, so much more intuitive compared to TensorFlow. It felt super natural when building neural networks with PyTorch
what kind of IDE is this at 21:57 ? using an unregistered version at AMD 😮
This is an extremely well-made documentary
Very inspiring! Thank you for the great project! I have been using it a lot.
Forgot the time stamp, but there's a wrong caption typed as "Stable" when they said "Stability (AI)".
Pleasing to watch, just wish there was title and company of the various persons, i found myself search them constantly to give me context
Wow, this explains a lot about the open source nature of meta…. They need AI to scale so they can scale. Interesting!
this documentary is clear evidence of how communistic use of means of production to share innovations work best than capitalistic competition trying to destroy the adversary in the market
People always say that China is way ahead of the US in AI, but perception truly isn't reality.
@17:00 do u really need the zipper visuals to make the point?
Please don’t bake the captions into the video. If I cannot understand them, please put the transcription into the toggled closed captions. It’s distracting when not necessary like when someone is actually speaking a different language than the rest of the video (English in this case).
skill issue
12:28 isn't that GANs are short for Generative Adversarial Networks, chat am i wrong?
correct haha i made the same comment just now
I think the entire point is missed.
The acceleration that LLMs offer is the ability to access and distill information at an unprecedented rate
wait not the official Pytorch Document but Documentary????? Mannnnn as a AI researcher a big thank you :):):)
what kinda mouse is that that Yangqing is using?! Never seen it before
Anyone reading this had better never give up trading.
Everything should be open source. Innovation slows down as soon as the guys in the suits come in with their whips and start the endless pursuit of returns stressing everyone out.
What’s that monitor they’re programming on?
Pytorch truly brought modularity in AI and built a reproducible ecosystem.
Thank you.
Someone please explain to me what AMD is doing in this documentary, and why there is no mention of NVIDIA?
William Falcon made a mistake? GAN's = Generative Adversarial Networks
not Generative Auto Encoders?
Now let's just hope it doesn't make us unemployed and turn the world in chaos :)
It's inevitable. The pandora's box is opened.
But also I think and believe, almost 90%, the future would be good and better than current one (if many companies can manufacture silicon chips and there are open source models).
Great documentary!
where are videos about Nvidia's contributions?
Nvidia couldn't hear the request for participation over all that money they're making. 💰💵💵💵
@@darylallen2485 ABSOLUTELY! Jensen for the win!!
Microsoft and Nvidia are at the forefront of the AI revolution. SmythOS can help you tap into this power to drive innovation in your business. #AI #SmythOS
God I love this documentary !
They still use NCHW instead of NHWC.
Thank you for this.
Даже если связку закроют - я могу быть рад, что успел 3 раза и все, спокойно себе больше не переживаю, заработал без азарта, но так даже лучше)
brings back memories from Facebook AI days
pretty cool, although i prefer tensorflow.
I love PyTorch
cool documentary guys!!
Oh! Why didn’t Nvidia sponsor? They are not proponents of Open Source?
shots fired at tensorflow
awesome editing.
I've looking for a good documentary, maybe I should do indie or corporate documentaries
scikit-learn is the og ML library, dont change my mind
Amazing!!!
12:29 Generative autoencoders?????
GAN does not stand for generative autoencoders but generative adversery network
lets goo !
World experts on training models.
ptrblck the legend!
Technology goes too fast for me to keep up. I know some Python and now I just learn PyTorch exist and it’s the next BIG thing. I have the impression that only the first people who created it and use it may have some future from it.
Wish I could ask for an internship just like that. Have not had any luck this whole year. Mostly rejections.
i love my pytorch!
Nothing you can't do with numpy, a cpu and a couple of years give or take.
can you give a good example in support of your statement ? Seems pretty luring.
@@ashutoshpadhi2782 I mean it was just a joke. Numpy and pytorch are similar in certain ways like handling tensor manipulation but pytorch is optimized for machine learning with gpu functionality, gradient descent and other things built in. So by all means reinvent the wheel with numpy but as I said even if you did it would take a lot longer to compute than with pytorch, lol.
Sure... just list the names of the speakers instead of including their job titles.
Lol