Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

GPTs are generating sequences in a left-to-right order. Is there another way? With François Fleuret and @evanncourdier, in partnership with SkySoft-ATM, we developed σ-GPT, capable of generating sequences in any order chosen dynamically at inference time. 1/6

379,989 görüntüleme • 2 yıl önce •via X (Twitter)

10 Yorum

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

We only need to shuffle the input sequence and to use a double positional encoding, so that the model is informed of what token to generate at each position of the output sequence. 2/6

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

This opens a multitude of possibilities. Such as: (A) Conditioning on any subset of tokens from a sequence or (B) Estimating the probability of the remaining tokens at once. 3/6

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

(A) is done by putting the conditioning tokens “first”, and (B) is done by putting all the remaining tokens as “the next”. The latter requires carving the attention matrix. 4/6

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

We combine these two functionalities into a burst-sampling scheme with rejection. It generates all the remaining tokens in one shot, and keeps only the ones that are self-consistent. This allows sublinear sequence generation similar to diffusion. 5/6

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

With the advantages that simple tasks can be fully completed in a few steps. 6/6 (with a diffusion comparison)

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

More detail about this work in our paper, that was just accepted at @ECMLPKDD. Paper: Demo: Code: in preparation.

Jason Stockwell profil fotoğrafı
Jason Stockwell2 yıl önce

@francoisfleuret @EvannCourdier @SkysoftATM so the first words that come out are the most important and limit the following possibilities. cool.

Arnaud Pannatier profil fotoğrafı
Arnaud Pannatier2 yıl önce

@francoisfleuret @EvannCourdier @SkysoftATM I agree that it can be limiting in a way, but it can also be useful for the model: fixing some tokens might give the overall structure of the text it want to generate, and then it only has to fill in between.

Rob Leclerc profil fotoğrafı
Rob Leclerc2 yıl önce

@francoisfleuret @EvannCourdier @SkysoftATM We think multidimensionally, but we communicate serially. Intuitively this feels like it has legs.

liu grey profil fotoğrafı
liu grey2 yıl önce

@francoisfleuret @EvannCourdier @SkysoftATM I'm sorry what? Half an hour ago I was tweeting about an architecture just like this combined with type checking/linting information that is tokenized and injected to improve code generation.. And you just post a paper on the difficult part of that project? Just.. right now? O.o

Benzer Videolar