Тёмный

Machine Learning From Zero to GPT in 40 Minute 

Brainxyz
Подписаться 23 тыс.
Просмотров 28 тыс.
50% 1

Source Code: github.com/hunar4321/reweight...
Title: Fast Paced, Programing Tutorial on Deep Neural Networks From the Basics To GPT and Beyond
Topics:
00:00 Machine Learning
05:12 Evolutionary Algorithms
08:59 Gradient Descent
11:38 Neural Networks
13:12 Deep Learning
17:16 PyTorch
23:56 Language Modelling
29:51 Convolutional Neural Networks
36:28 LSTM - Attention - Transformer
41:35 Re-weight GPT
44:59 A.I. alignment
Website: www.brainxyz.com/
#pytorch #python #chatgpt #deeplearning
Disclaimer: All materials in this video are used for educational purposes and fall within the guidelines of fair use. No copyright infringement is intended. If you are or represent the copyright owner of materials used in this video and have a problem with the use of said material, please contact us. Thanks
Walk-through coding tutorial on machine learning from basics to gpt
Tags:
#chatgpt #artificialintelligence #pytorch #tutorial #dalle3 #agi #grok
pytorch python numpy machine learning neural network brainxyz brain coding intelligence self-attention transformers transformer architecture deep learning data science data analysis linear regression bard ai google bing openai diffusion model generative ai agi artificial general intelligence Gemini chatgpt gpt gpt-3 gpt-4 gpt-5 sora Grok large language models LLMs dalle3 Q-learning Q* mamba state space models

Наука

Опубликовано:

 

2 июн 2024

Поделиться:

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист
Посмотреть позже
Комментарии : 62   
@brainxyz
@brainxyz Год назад
Please checkout the source code in the description. Some people like fast paced videos others like slow pacing but the best way to learn is to do it yourself. Pause the video at each step and follow the instructions. If you stuck at any point, compare your code with the source code provided in "tutorials" folder. The GitHub link is in the description. Best wishes
@nicdemai
@nicdemai Год назад
Thank you for this video!
@nevinb60
@nevinb60 Год назад
You did such a great job! If you had a series like this I would definitely watch
@Octo_Fractalis
@Octo_Fractalis Год назад
👏, very good, can you mix fractals, AI and artificial life?
@calabisan
@calabisan Год назад
I'm speechless. No one else has managed to sprint from light switches to the leading edge of technology in less than an hour, creating the impression as if it were a leisurely walk - very paradoxical. Overall, it was incredibly well done! Many thanks!
@joey199412
@joey199412 Год назад
This has been BY FAR the best machine learning video I've ever seen. How you went through different systems and not only show how they work in code but also go through the upgrades and how they relate to other subsequent inventions shows us how these concepts evolved, how they relate to math, how to intuitively understand them and how to implement them. All in just under 50 minutes. This video is a masterpiece.
@corallaroc2946
@corallaroc2946 Год назад
You cannot make AI any more engaging. Thank you for this. If you write a book it will become a bestseller.
@Santzes
@Santzes Год назад
I've only watched like 20 minutes but I can already say this is the best video about neural networks I've ever seen. I've already understood many things I've almost understood but never clicked before
@kieuphongeg
@kieuphongeg Год назад
I'm too, easy to know neural network
@agentzero3209
@agentzero3209 Год назад
This is a good example of free high quality educational content available for free online.
@blutoo1363
@blutoo1363 Год назад
This video is the fucking bomb. Crisp editing, top notch humour and clear understanding. Hope you continue to bless us.
@user-me9em9fy4k
@user-me9em9fy4k Год назад
You've successfully helped me break through a mental wall and I cannot thank you enough.
@daianag8457
@daianag8457 4 месяца назад
You know... I have to get up in 3 hours to go to work.... and yet here I am. Great channel
@The-Martian73
@The-Martian73 Год назад
Please Hunar keep doing this, and do never stop ... your videos are legendary and iconic .
@liquidvicinity
@liquidvicinity Год назад
I always get super excited when I see a brainxyz video in my notifications
@norexmusic7888
@norexmusic7888 Год назад
I've been waiting for this!
@movAX13h
@movAX13h Год назад
Excellent! Thank you very much for this detailed video.
@nevereveraskedforthis
@nevereveraskedforthis Год назад
That is an involvement!
@coolsai
@coolsai Год назад
Great video, This is what i need
@madankd
@madankd 11 месяцев назад
Amazing content man it's gold
@sukhWins
@sukhWins Год назад
Thanks this is great knowledge. Although it is fast paced but listening it on repeat you can gradually understand it
@aleksszukovskis2074
@aleksszukovskis2074 3 месяца назад
this is good and beginner friendly. youve earned a sub
@hassanmehmood8934
@hassanmehmood8934 9 месяцев назад
Thanks for creating this wonder full explanation. Best video i ever seen on youtube. Thanks @brainxyz superb content with great explanation. kindly create more videos like this.
@hawrotaha2647
@hawrotaha2647 Год назад
May God bless you Keep up bro
@chelsyma3926
@chelsyma3926 5 месяцев назад
Please keep producing more videos like this!
@VenturaPiano
@VenturaPiano 3 месяца назад
I rarely comment on videos, but I have to say, this information is stellar!
@jpr9734
@jpr9734 Год назад
very good!!!
@ThinAirElon
@ThinAirElon 6 месяцев назад
Awesome ! what software did you use for animations ?
@mbrochh82
@mbrochh82 Год назад
This might be the Transformer video to end all Transformer videos...
@generationgap416
@generationgap416 3 месяца назад
Are you saying this might transform the Transformer game? Next, you will be telling us that attention is all we need. Or we should use our multiple heads?
@waze84
@waze84 Год назад
Good work man, can you make a more elaborate tutorial with more example in short video playlist, that will be amazing
@ArvindDevaraj1
@ArvindDevaraj1 3 месяца назад
mind blowing explanation
@oooo0oooo0oooo
@oooo0oooo0oooo Год назад
wow this was information dense and fun to follow. seems like some influence from Schmidhuber during the last few minutes :)?
@SiEmG
@SiEmG Год назад
true
@aleksszukovskis2074
@aleksszukovskis2074 3 месяца назад
12:19 there is a mistake in what youre saying. "input of previous layer will become the output of the next layer" but its actually "output of previous layer is input of the next layer"
@bartoszstyperek6306
@bartoszstyperek6306 Год назад
I do not get the 34:14 part. You said we have to 'pass inputs with various context lengths' , but you changed the sampled output ys length, from [i+ins+1:i+ins+1] to [i+1:i+ins+1] range : ys = torch.stack([data[i+1:i+ins+1] for i in b]) . so from [cat] -> [s] we got to [cat] -> [ats] . I could use some extra explanation, what is going on here .
@brainxyz
@brainxyz Год назад
Yes: "from [cat] -> [s] we got to [cat] -> [ats]" The output now is a shifted version of the input by one letter. If you train now, you'll use the input [cat] to predict each of the letters in [ats] as follows: [cat]->[a] [cat]->[t] [cat]->[s] To pass the inputs [cat] with various context lengths, you need the dot product between the inputs and a matrix of 1s masked with 0s as explained from 34:30 onward. Then, you'll get: [c]->[a] [ca]->[t] [cat]->[s]
@bartoszstyperek6306
@bartoszstyperek6306 Год назад
​@@brainxyz now I get it (was not sure about the tri matrix, but now I do) . This is best video on subject. Even easier to follow than karpathy tutorial (at the end he rushed the whole q,k,v stuff) .
@starplatinum3305
@starplatinum3305 6 месяцев назад
I HAVE TO SUB TO THIS DUDE, THIS DUDE IS CRAZY 😧
@HemnAhmedhm
@HemnAhmedhm Год назад
Excellent
@imaspacecreature
@imaspacecreature Месяц назад
Wonderful content
@skilz8098
@skilz8098 Год назад
From a programming and engineering perspective I do find these topics quite interesting. However, from a consumer standpoint where the use of GPUs became a viable option in both A.I. programming as well as with Bitcoin Mining... these scalpers are responsible for driving the cost of GPUs to an all time high causing them to be ridiculously overpriced for the conventional gaming markets and causing them to be almost unaffordable for the basic consumer. Instead of high end GPUs costing a moderate few hundred dollars, they're now in the thousands. This observational rant has nothing to do with this videos quality content and presentation. This video is actually quite good and the hard work, dedication and time invested is much appreciated, and it was quite enjoyable. It's just that due to these miners and large corporations buying up all of the GPUs on launch in mass bulk causing GPU availability to become scarce and almost unaffordable the GPU manufacturers have or are taking advantage of this by raising their MSRP values keeping them at an all time high without even considering to lower them making them the new standard MSRP. So instead of buying a top of the line GPU for about $500, a mid level GPU for about $200-300 and an entry level GPU for about $100... your newer entry level GPUs are closer to about $300-500, your mid level are now about $500 - $800, and your top of the line GPUs are now over $1k and some are even as high as $2k. GPUs now cost almost as much as the rest of the entire system. A good high end gaming rig use to only cost about $1,200 - $2,000 and now they're closer to $3-4k or even more. How can we resolve this current issue? We need manufacturers to make dedicated cards that are designed specifically for these task. We need dedicated cards just for A.I. systems, and dedicated cards just for Mining. And let the miners, and large corporations to foot the bill for these more expensive and dedicated task cards while leaving the Gaming Markets and their GPU prices and availability alone. These current GPU prices (MSRPs) need to be dropped by either 1/2 or even 2/3rds of their current MSRP values! Just an astute observation from a consumer in disgust. As for the material of this video and the content quality of it and the engineer within, excellent job sir! Keep up the good work!
@generationgap416
@generationgap416 3 месяца назад
One Answer is Sparsity and Quantization
@galbalandroid
@galbalandroid Год назад
Thanks for this amazing video! I gonna be watching it few more times just to make sure I understand the things good enough! What is the code editor you're using?
@brainxyz
@brainxyz Год назад
Spyder
@muajkoob-db7qs
@muajkoob-db7qs Год назад
A for alpa in greek term
@ojuliomkt
@ojuliomkt Год назад
good
@amparoconsuelo9451
@amparoconsuelo9451 9 месяцев назад
If I will order from Amazon a GPT assembly kit, what would it deliver me? How much would the kit cost?
@kruptworld
@kruptworld Год назад
im stuck at the "finding solution" part. my error doesn't seem to be able to pass 10.01. also what tool did you use to make the plot interactive?
@brainxyz
@brainxyz Год назад
Compare your code with the source code, you may find where is the mismatch (GitHub code link is in the description section)
@IronMechanic7110
@IronMechanic7110 Год назад
@@brainxyz Is it possible to create a chatbot with this ??
@brainxyz
@brainxyz Год назад
@@IronMechanic7110 Yes, but your model size should be very large and you have to train it on a very large corpus of chat & conversation texts.
@IronMechanic7110
@IronMechanic7110 Год назад
@@brainxyz You shoud make a tuto about it. Need maybe mutliples of gpu.
@flaiskai2371
@flaiskai2371 Год назад
ياريت تعملو نفس الفيديوات بالعربي ......مشكورين
@MrMehrd
@MrMehrd 11 месяцев назад
Thx
@wilfredomartel7781
@wilfredomartel7781 5 месяцев назад
Could you include in your list Mamba?
@_belloul6706
@_belloul6706 5 месяцев назад
Shameful DNA was being this for billion years, this passage made me fall from my chair 😆
@matiabem2346
@matiabem2346 Год назад
itsbetter explanation 10q
@andrewdunbar828
@andrewdunbar828 Год назад
One alphabet ought to be enough for anybody.
@yudelko
@yudelko 7 месяцев назад
what is the econmical return(gain) of building gpt from scratch?
@fcf8269
@fcf8269 Год назад
Great video, but the tone and the pacing is what kill it
@mansursaz1421
@mansursaz1421 Год назад
good
Далее
ChatGPT: 30 Year History | How AI Learned to Talk
26:55
The Most Important Algorithm in Machine Learning
40:08
Просмотров 233 тыс.
МАЛОЙ ШАНТАЖИСТ
00:34
Просмотров 140 тыс.
Is it Cake or Fake ? 🍰
00:53
Просмотров 1,7 млн
New Gadgets! Bycycle 4.0 🚲 #shorts
00:14
Просмотров 10 млн
Watching Neural Networks Learn
25:28
Просмотров 1,1 млн
Building a GPT from scratch using PyTorch - dummyGPT
11:23
Why Neural Networks can learn (almost) anything
10:30
Индуктивность и дроссель.
1:00
iphone fold ? #spongebob #spongebobsquarepants
0:15
Просмотров 345 тыс.