Загрузка видео...

Не удалось загрузить видео

На главную

Just found the dLLM library to create Diffusion Language Models It's still early but it's insanely fun to experiment with diffusion (training, inference, eval) dLLM has the potential of becoming the main library for diffusion LLMs

75,593 просмотров • 1 год назад •via X (Twitter)

Комментарии: 22

Фото профиля Maxime Labonne
Maxime Labonne1 год назад

💻 Library: Source: by Individual-Ninja-141

Фото профиля truth.phd
truth.phd1 год назад

Diffusion LLMs are like teaching language models to dream before they talk. Instead of guessing the next word, they refine whole thoughts from noise, piece by piece. dLLM might just make text generation feel more like painting sentences than typing them. It’s chaos turning into poetry, one token at a time.

Фото профиля Seba
Seba1 год назад

liquid training a ?

Фото профиля Sciumo
Sciumo1 год назад

Can these models be distilled or merged?

Фото профиля Maxime Labonne
Maxime Labonne1 год назад

Definitely! Especially merging

Фото профиля Sciumo
Sciumo1 год назад

I suspect there will be some fantastic results, especially with reasoning isolation, distillation and transfer.

Фото профиля Sardor Madaminov
Sardor Madaminov1 год назад

Dang, UI is well put together in the terminal and it looks so fun to try this model

Фото профиля Klemen Simonic
Klemen Simonic1 год назад

We’ve been experimenting with it for speech understanding too; the idea of generating or reasoning over continuous representations instead of discrete tokens feels like the next frontier.

Фото профиля Enohsla
Enohsla1 год назад

Have to try it out soon.

Фото профиля Sir Mr Meow Meow
Sir Mr Meow Meow1 год назад

neat

Фото профиля Lyceum
Lyceum1 год назад

Sounds like a wild ride!

Фото профиля Omo Ìyá Àgbà
Omo Ìyá Àgbà1 год назад

I did something like this with a critique that self corrects it self also. This one is equivalent to autoregression

Фото профиля Roshx
Roshx1 год назад

So cool

Фото профиля X. Dong
X. Dong1 год назад

A model can think in Diffusion and talk in Autoregression.

Фото профиля Ditectrev
Ditectrev1 год назад

Nice!

Фото профиля Thanh Doan
Thanh Doan1 год назад

curious to see what kind of unique applications this enables compared to standard transformer architectures

Фото профиля IkhlasulHanif
IkhlasulHanif1 год назад

@rdiandaru

Фото профиля Himanshu Kumar
Himanshu Kumar1 год назад

That's a cool find, Maxime! The potential of dLLM sounds quite exciting, especially for diffusion LLMs, don't you think?

Фото профиля Tyler Wolfe
Tyler Wolfe1 год назад

Very cool

Фото профиля Ansar Ullah Anas
Ansar Ullah Anas1 год назад

The idea of combining diffusion models with LLMs seems like it could open up new research directions. How beginner-friendly is dLLM for someone new to this?

Фото профиля Sense
Sense1 год назад

This is new to me. Can this generate tokens in parallel or is it still linear?

Фото профиля Hare Fu
Hare Fu1 год назад

I suppose dLLM is similar to BERT

Похожие видео