[ 100k Special ] Transformers: Zero to Hero

Let's talk about transformers from scratch.

0:00 Thank you for 100K!
0:29 Transformer Overview
12:27 Self Attention
26:40 Multihead Attention
39:31 Position Encoding
48:51 Layer Normalization
1:00:50 Architecture Deep Dive
1:27:56 Encoder Code
2:16:10 Decoder Code
2:54:11 Sentence Tokenization
3:12:26Training and Inference

MATH COURSES (7 day free trial)

OTHER RELATED COURSES (7 day free trial)

#chatgpt #deeplearning #machinelearning #bert #gpt

[ahmadfaraz]

This was the best playlist on YouTube on Transformers

[ShakirulHasanShahad]

literally i've searched so many articles blogs and youtube videos to understand transformer from the core basic. And this video come up with all the stuffs, and i must say best video everr!

[savindudulanakarajapaksha]

Best video to learn about transformers. Since you are going through same theories again and again more I heard them more I understood them. Thank you very much.

[EpicGamer-uxtu]

Onwards and upwards to you too Ajay, great vid mate, I actually watched the whole thing

[SarathKumar-gv]

Hey Ajay, I've watched numerous videos on transformers, but yours stands out as the best—super clear and easy to grasp! Your explanations are fantastic. Could you consider covering transformers for time series in your future videos? That'd be incredibly helpful! Thanks for the great content!

[yaswanthyalamuri]

thank you so much ajay, your channel is one of the best in youtube for learning ml

[manfredgriesser]

I would have never understood that topic without your video. Thank you.

[jacehua]

Been following you for a while. Congrats! You definitely made my NLP journey more enjoyable with your informative and entertaining videos.

[chinton]

This is the best playlist ever. Thank you.

[vishnuthirtha]

Amazing content, you made NLP naturally processable for me.

[rakshitaralimatti]

The Best explanation on transfer with code, even tough I am a AI developer and know these things pretty well I watched the full video
Love from Banglore

[mamoun]

Thank you for sharing your knowledge with us. This is pure gold !

[acaudio]

I think at this point I represent a heavy percentage of your views. Great work bro.

[pablovaras]

the best transformer video i have watched

[chrisogonas]

This is incredible and well illustrated. Thanks for putting it together.

[thvk]

I have seen so many transformers videos but this one is outstanding, I also want to request you to make a video on vision transformers too❤

[sumitsp]

27:40 isn't the key, query and value vectors computed using weight matrices for each of them which have learnable parameters. The input vector is multiplied by weight matrix corresponding to each Q, V and K.

[ramanShariati]

best video on transformers ever. even better than Andrej Karpathy's video. you are awesome bro. keep up the great work.

[hazardousharmonies]

Hi Ajay: I have a question on your excellent video. At 10:28 (and also earlier) you say the 2nd MultiHeadAttention block in the Decoder outputs attention vectors **for each english AND french word** . Is this correct? I think the input to the Linear layer can be **only french words** (and not english). So is a) the input to the decoder's FeedForward **both English and French but its output French**, or b) the input to the decoder's FeedForward only French words, c) something else I have not yet grasped? Thanks again for this great video :)

[nguyennhi]

omg, you're deserve more and more subscribers