The slides are posted here: cs.uwaterloo.ca/~ppoupart/teaching/cs480-spring19/schedule.html
@tarunluthrabk
3 жыл бұрын
Hello Professor. Your explanations are amazing. Kindly pin this comment or add in description so that it is visible to everyone.
@majid0912
Жыл бұрын
ّDear Pascal, I was wondering if you have any presentations to describe the article titled Neural Machine Translation By Jointly Learning To Align And Translate.
@SuperOnlyP
3 жыл бұрын
Finally, Someone can explain simply what queries, keys , values are in the transformer model . Thank you Sir !!!
@mathisve
3 жыл бұрын
Yeah, I don't understand why nobody else goes over this seemingly pretty important detail
@stackoverflow8260
3 жыл бұрын
Wow, I was gonna ask, why didn't he explain or give an example for query, key, value in the case of a simple language translation or modelling example. Machine learning community is not very good at conveying their ideas, when you can't put stuff in rigorous mathematics at least use a lot of pictures and many examples at every possible step.
@andrii5054
3 жыл бұрын
I can also recommend this explanation: kzitem.info/news/bejne/zoOXmISjknyddqQ It has helped me a lot
@SuperOnlyP
3 жыл бұрын
@@andrii5054 The video really simplify the concept. Thanks for sharing !
dont worry, im sure the next visit to a public internet forum will once again obliterate hope in humanity
@100vivasvan
3 жыл бұрын
haha same here
@dilettante9576
Жыл бұрын
Cured my ADHD
@Mrduralexx
Жыл бұрын
This video gave me depression…
@UmerBashir
Жыл бұрын
@@Mrduralexx yeah its a different level of anxiety that it instills
@sudhirghandikota1382
4 жыл бұрын
Thank you very much Dr. Poupart. This is the best explanation of transformers I have come across on the internet
@GotUpLateWithMoon
3 жыл бұрын
This is the best lecture on attention mechanism I can find! Thank you Dr. Poupart! finally all the details made sense to me.
@cwtan501
3 жыл бұрын
By far the best I have seen to explain multiheaded attention
@AI_ML_iQ
Жыл бұрын
In recent work, titled "Generalized Attention Mechanism and Relative Position for Transformer" , on transformer it is shown that different matrices for query and key are not required for attention mechanism in Transformer thus reducing number of parameters to be trained for Transformer of GPT, other language models and Transformers for images/videos.
@dennishuang3498
2 жыл бұрын
Very enjoyed your lecture, Professor Poupart! Very informative and simplified many complicated concepts. Thank you very much!
@graceln2480
2 жыл бұрын
One of the best explanations for attention & transformers in KZitem. Most of the other videos are junk with authors pretending to understand the concepts and just adding to the KZitem clutter.
@momusi.9819
4 жыл бұрын
Thank you very much, this was by far the best explanation of Transformers that I found online!
@dansplain2393
2 жыл бұрын
I was about to type the same.
@moustafa_shomer
2 жыл бұрын
This is the best Transformer / Attention Explaination ever.. Thank you
@hamzaaliimran6441
6 ай бұрын
one of the best and detailed lecture on attention on youtube I must say.
@JMRC
4 жыл бұрын
Thank you to the person asking the question at 28:49! The softmax gave it away, but I wasn't sure.
@drdr3496
Жыл бұрын
This is the single best video on "Attention is all you need", attention, transformers, etc. on the Internet. It's simple as that. Thanks Dr Poupart.
@bleacherz7503
Жыл бұрын
Why does a dot product correlate to attention?
@drdr3496
Жыл бұрын
@@bleacherz7503 a dot product between two vectors shows how similar they are
@seldan6698
Жыл бұрын
@@drdr3496 nice. Can you explain me whole query , key and value process for some example like " the cat sat on the mat". What is query, key and values for this sentence
@robn2497
3 ай бұрын
ty
@utkarshgupta7364
3 жыл бұрын
Most awesome video on transformers one could find on youtube
@sajalvasal5073
3 жыл бұрын
You are a blessing, finally understood a very important concept.
@aadeshingle7593
9 ай бұрын
Thanks a lot Professor Poupart one of the best explanation for maths behind transformers!
@TylerMosaic
3 жыл бұрын
wow! love the way he answers that great question at around 50:52 : “why we dont we implement the mask with hadamard product outside of the softmax?”. brilliant prof.
@richard126wfr
2 жыл бұрын
The best explanation of attention mechanism I found on KZitem is the making pizza analogue by Alfredo Canziani.
@mi9807
10 ай бұрын
One of the best videos!
@benjamindeporte3806
7 ай бұрын
I eventually understood the Q,K,V in attention. Many thanks.
@fengxie4762
3 жыл бұрын
A great lecture! Highly recommended!
@underlecht
3 жыл бұрын
I would call this the best explanation of attention/transformers on youtube i have found so far.
@vihaanrajput8082
2 жыл бұрын
His toturial video is my favorite timepass, specially at night,Hail to prof. Poupart
@orhan4876
6 ай бұрын
thank you for being so thorough!
@SivaKumar-gs5ku
2 жыл бұрын
This is the best video Internet about Transformers network
@ghostoftsushimaps4150
9 ай бұрын
Bhaiji love from India. Is lecture ko araam se dekhunga
@weichen1
4 жыл бұрын
I am not able to find a better video than this one explaining attention and transformer on the internet
@MrFreemindonly
3 жыл бұрын
I totally agree, He is genius!
@xhulioxhelilai9346
2 ай бұрын
Thank you for the very comprehensive and understandable course. Being in 2024 I can say that I can understand even better and easier this course using GPT-4.
@sandipbnvnhjv
Жыл бұрын
I asked chatGPT for the best video on Attention and it brought me here
@ibrahimkaibi4200
3 жыл бұрын
A very interesting explanation (wonderful)
@jinyang4796
3 жыл бұрын
Thank you for the clear explanation and well-illustrated examples!
@giorgioregni2639
3 жыл бұрын
Best explanation of transformer I have ever seen, thank you Dr Poupart
@parmidagranfar4861
2 жыл бұрын
finallu understood what is going on . most of the videos are so simple and skipped math . i liked it
@jelenajokic9184
2 жыл бұрын
The simplest explanation of attention, thanks a lot for sharing, great lectures🤗!
@aponom84
3 жыл бұрын
Nice lecture! Thanks!
@HeshamChannel
Жыл бұрын
Very good explanation. Thanks.
@syedhasany1809
4 жыл бұрын
This was a great lecture, thank you.
@firstnamelastname3106
3 жыл бұрын
thank you my man, u saved me
@pred9990
4 жыл бұрын
Cool lecture!
@benjaminw2194
2 жыл бұрын
I'm a novice and have been praying to get someone who discusses these papers. You're an answered prayer! Great lecturer.
@faatemehch96
2 жыл бұрын
thank you, the video is really useful. 👍🏻👍🏻
@minhajulhoque2113
Жыл бұрын
Great video!
@shifaspv2128
Жыл бұрын
Thank you so much, the brainstorming
@cedricmanouan2333
3 жыл бұрын
very interesting and useful. Thanks Sir
@aileensengupta
Жыл бұрын
Big fan, big fan Sir!! Finally understood this!
@ritik84629
Жыл бұрын
True
@Antony25rages
3 жыл бұрын
Thank you for this :)
@user-or7ji5hv8y
3 жыл бұрын
Does anyone know? Which NMT video has the previous intro to Attention the professor cites in this video? I couldn’t find his video on neural machine translation.
@seminkwak
3 жыл бұрын
Beautiful explanations
@brandonleesantos9383
Жыл бұрын
Truly fantastic wow
@MustafaQamarudDin
4 жыл бұрын
Thank you very much. It is very detailed and captures the intuition.
@syphiliticpangloss
4 жыл бұрын
Could you explain what the model class looks like then? What is the capacity? What is the "unconstrained" version with higher capacity? I was full statistical learning theory style discussion in all pedalogical discussions. I don't understand how people think they understand this. If your life depended on it, would you feel confident in recommending one of these setups? What questions would you have to ask about the data, the model architecture, the observation process? You need worst case bounds, model complexity etc. I see none of that here.
@1Kapachow1
3 жыл бұрын
@@syphiliticpangloss Well, in deep learning the theory is far behind engineering. When people say they understand this lecture, they don't mean worst case bounds (which I strongly doubt anyone in the world knows how to calculate for this, without adding so many relaxation assumptions which make it basically irrelevant, like convexity etc.), they just mean that: 1. Engineering wise they understand how to build and use it 2. They feel they grasp enough intuition to what is the purpose of each sub-block and why it was added. I don't think anyone truly "understands" much simpler models in DL than transformers, which perform in a far superior level to classical machine learning methods. For example, fully convolutional neural networks, trained with Adam optimizer, based on back-propagation, using BN.
@syphiliticpangloss
3 жыл бұрын
@@1Kapachow1 So can someone explain what the transformer is doing then in a precise way? I would accept answers that reference probability distributions and predictive goals or computation description of components like NAND gates etc. Also accepted would be anything related to the eigenvalues, stability, curvature etc. There are lots of people trying to talk about this stuff. For example arxiv.org/abs/2004.09280 Or Vapnick. To be perfectly clear, I think today we tend to say there are only two things really: a) "data" i.e. observations usually dozens to millions from some process we take to be slowly changing at most and b) predicates/models/architecture/constraints ... "observations" usually less than dozens, usually manually constructure (from other experiments and observations sets perhaps). To each of these we usually have some sort of "narrative" about where each came from, a way of describing it in some way to humans. The second thing is what I'm getting at. "Architecture" is a model constraint. If it is just pulled from thin air without undestanding the problem, the meta-problem etc, it is quite likely that there are buried problems, secret reasons for architecture choices that are not being disclosed or realised. Getting better at describing these models/arch/predicates is how we progress.
@hariaakashk6161
3 жыл бұрын
Great explanation sir... Thank You! Please post more such lectures and I would be the first to look at it...
@fit_with_a_techie
2 жыл бұрын
Thank you Professor :)
@greyreynyn
3 жыл бұрын
41:14 Question, on the output side, why isn't there an additional feed-forward layer between the masked self attention in the output and the attention to the input? And maybe more broadly what are those feed forward units doing?
@gudepuvenkateswarlu5648
2 жыл бұрын
Excellent session....Tq professor
@evennot
4 жыл бұрын
19:00 it's basically an exclusionary perceptron layer, isn't it? (also could be called fuzzy LUT) I'm sure it was used before for the attention emulation
@akashpb4183
3 жыл бұрын
Beautifully explained.. things seem clear to me now .. Thanks a lot sir!
@c.l.1269
3 жыл бұрын
Great lecture! Thank you Professor!
@abhijeetnarharshettiwar6175
2 жыл бұрын
Thank you so much for great explanation, professor.
@Vartazian360
6 ай бұрын
Little did anyone know just how groundbreaking this foundation would be for Chat GPT / GPT 4.
@markphillip9950
3 жыл бұрын
Great lecture.
@sheikhjubair7133
4 жыл бұрын
Very clear explanation
@AnonTrash
Жыл бұрын
Beautiful.
@aymensekhri2133
Жыл бұрын
Thank you very much Sir!
@mohamedabbashedjazi493
3 жыл бұрын
Softmax is computationally expensive, I wonder if this can be replaced somehow with another function to produce probabilities since Softmax is present in many places in all the blocks of the transformer network.
@nafeesahmad9083
2 жыл бұрын
Woohoo... Thank you so much
@aricircle8040
Жыл бұрын
Thank you very much for sharing that great lecture! Shouldn't it be the attention vector instead of the value? at 27:44
@goldencircle4331
Жыл бұрын
Huge thanks for putting this online.
@sienloonglee4238
Жыл бұрын
very good video!😀
@weiyaox6896
2 жыл бұрын
Best explanation
@jaeyoungcheong1767
3 жыл бұрын
Clearly! Thanks
@yd42330
3 жыл бұрын
Question about positional encoding. If we sum the Word Embedding (WE) with the Positional Encoding (PE) how does the model tell the difference between WE = 0.5, PE = 0.2 and WE = 0.4 and PE = 0.3 ? (Different words that are at different positions can yield the same value) Why not keep the PE separate from WE?
@opencvitk
Жыл бұрын
the explanation of K,V and Q is great. unfortunately i lost him as soon as he started on multi-head. must be that the single head i possess is empty :-)
@greyreynyn
3 жыл бұрын
46:45 - Also, the output shape is the same as the input shape right? ie, the size of the input sequence?
@reuben3648
Жыл бұрын
Thank you soo much!!!
@yashrajwani3322
3 жыл бұрын
great explanation
@hackercop
2 жыл бұрын
This was a great lecture - really explained this to me thanks
@yen-linchen7398
Жыл бұрын
Thank you!
@444haluk
3 жыл бұрын
I heard queries, keys & values were primative concepts and counter-intuitive, but I didn't know it was THIS primative.
@soumyarooproy
4 жыл бұрын
Great point at 50:50 👍
@amitvikramsingh327
3 жыл бұрын
Thank You.
@kungchun9461
3 жыл бұрын
This year should be the "tranformer year" as there a breakout in domain of CV.
@justinkim2973
Жыл бұрын
Best video to watch on the first day of 2023
@autripat
3 жыл бұрын
At 1:18:22, the professor refers to BERT and a "Decoder transformer that predicts a missing word". To me, BERT is a masked Encoder (not decoder). After all, BERT stands for bidirectional *encoder* representation from transformers. It's minor (and doesn't subtract from this great presentation), but can anyone comment?
@abdelrahmanhammad1020
2 жыл бұрын
Great lecture. And I believe you are correct, it seems there is a typo here. I was questioning the same!
@varungoel185
3 жыл бұрын
Around @29:50 mark, he first mentions that the key vectors correspond to each output word, but the slide mentions input word. Could someone please clarify this?
@chakibchemso
Жыл бұрын
and thats how gpt was born my fellas
@mohamedabdo-dl9dd
3 жыл бұрын
thanks professor for easy explain ... can you share powerpoint with us ..
@alexanderblumin6659
2 жыл бұрын
Very intersting lecture. Something that is not totally clear on minute 46: these multihead presented intuitievely as explicit 3 various filterts as in cnn to produce 3 corresponding feature maps,but on previous part of lecture its being said that multi heads are stacked one after another to produce at first info from (word i,word j) and second pairs of that stuff i.e one is the input to another one. So how to understand it in the rightr way? Seems like on minute 46 the inputs to each of the linear are the same but on lecture part it looks like one is going after another and intuitevly the pair of pais and so one changes the ouput size.
@leoj5891
2 жыл бұрын
does this normalization layer matter in the inference stage?
@anatolicvs
2 жыл бұрын
Dear Prof. Dr. Poupart, do we have chance to have your presentation that used at the lecture of 19 please ?
@prof_shixo
4 жыл бұрын
Thanks for the nice lecture. I am still confused regarding how transformers model can replace RNNs or LSTMs for general sequence learning. The size of a sequence might be very lengthy in some applications rather than just a sentence (which can be designed to be fixed in length) so how to deal with this especially if we need to keep the complete sequence with us as there is no recurrence? If the answer is to divide the sequence, then how to link different chunks over time without a recurrence or a carry over? (Loops over time)
@JAKKOtutorials
4 жыл бұрын
transformers are able to "query the recurrencies", think of it as instead of repeating the operation as in RNNs you just query 'x' times a database of the possible values and its given inputs and check if it matches the requirements, and because it's not a recurrence, repetition, you can make multiple of these queries, each being a new operation, at the same time!! each operation can be resolved without interference creating new tokens, or pieces, which represent convergence points in the data universe you are travelling. it's a huge improvement.. confirmed by the models shown at the end of the lecture. hope this helps :)
@venkateshdas5422
4 жыл бұрын
As JAKKO mentioned the transforms use the attention mechanism in a very efficient manner. The size of the sequence can be sufficiently longer than a sentence and still the attention mechanism will be able to capture the dependencies between the words at different positions. And this creates an efficient contextual representation of the sequence better than the normal input embedding vector. And this is how the complete input sequence is captured by the model without the recurrence. This is really a beautiful approach. (personal opinion)
@LironCohenProfile
3 жыл бұрын
At 31:20, shouldn't the last line be v_j (rather than h_j)?
@greyreynyn
3 жыл бұрын
45:50 - For the multiple linear transformations, are we applying the same linear transform to each set of Q/K/V in a "head" ? Or does each Q/K/V get its own unique linear transform applied?
@knoxvoxx
3 жыл бұрын
Unique linear transform each time I guess.( In the original paper, under section 3.2.2 they mention that " h times with different learned linear projections to dk, dk and dv respectively") If we take repeat scaled dot product attention 3 times , then we will have total of 9 linear projections.
@ryanwhite7401
2 жыл бұрын
They each get their own learned parameters.
@abhishekrohra9457
2 жыл бұрын
Good explanation
@greyreynyn
3 жыл бұрын
57:30 - I understand the normalization, but what's the intuition for adding? Does it just strengthen the signal from the input sequence?
@Victor-oc1ly
2 жыл бұрын
I can comprehend the additive part as incrementing the knowledge from the self-attention operation to the query itself. You can think of it as the query (let's call it x) is a proper question and the sublayer (multihead attention) operation output is the contribution to your inquiry from the oracle (the sub-layer before the feed-forward step), which can then be used to refine your question (the query). If you look at the paper, it's really LayerNorm(x + Sublayer(x)) what the authors write for the Encoder and Decoder stacks.
@ephremtadesse3195
2 жыл бұрын
Very helpful
@ryandruckman999
4 жыл бұрын
1:02:00 For the positional embedding, I am confused... It seems like the formula produces a scalar output (ie you put in a position in your sequence, you get out some sin/cos value). How does it become a vector?
@ryandruckman999
4 жыл бұрын
You could take the value for each dimension in your embedding and then you have a vector of values. But then it seems you'd be encoding the same information in every input, which doesn't seem helpful?
@venkateshdas5422
4 жыл бұрын
@@ryandruckman999 Actually while doing that the values fed into the cos and sin function also takes the scalar position of the word. So each positional embedding vector will have different values which in turn captures the position / order of the sequence. refer this article : kazemnejad.com/blog/transformer_architecture_positional_encoding/
@hemanthsharma5630
4 жыл бұрын
Saviour!!!
@samson6707
2 ай бұрын
46:11 i dont understand how there can be a concatention of the outputs followed by a linear combination. in my mind it doesnt make sense to do both. either the outputs are concatenated or added up in a linear combination but both...?
Пікірлер: 215