Hugging Face's logo
Yatharth Sharma's picture
320 365 610

Yatharth Sharma

YaTharThShaRma999
couragestrong's profile picture Jongsim's profile picture thomaskalnik's profile picture
Ā·

AI & ML interests

None yet

Recent Activity

updated a model 1 day ago
YaTharThShaRma999/voices
reacted to Ruurd's post with šŸ”„ 2 days ago
The past year I have been trying to get diffusion models to work for language generation, without having to retrain a LLM from scratch. And recently, we finally succeeded: We introduce "LAD: LoRA-Adapted Denoiser", a method to convert a LLaMA model into a text diffusion model using LoRA finetuning and structured input corruption. šŸŽÆ Try the demo and read the write-up here! https://ruurdkuiper.github.io/tini-lad/ Unlike autoregressive (word-for-word) models like ChatGPT, diffusion models iteratively refine a noised sequence. However, most current diffusion approaches rely on all-parameter retraining and repeatedly remasking tokens, which is costly and slow during both training and inference! šŸ§  With LAD: - We can finetune an autoregressive model for diffusive generation in just 10 hours on a single GPU. - Test-time compute is fully adjustable: fewer steps means faster outputs while more steps improve output quality. - Due to our unique noising schedule, remasking is not always needed during inference. All tokens are attended to in each iteration! šŸ” LAD is built using: ā€“ A frozen LLaMA-8B backbone ā€“ Structured noising: token swaps, duplications, replacements, span shifts ā€“ Modified attention masks for bidirectional decoding šŸ’” We show that even small, fast-trained models can perform diffusive generation ā€” with competitive benchmark performance, perplexity and more flexible test-time behavior than traditional transformers.
liked a model 3 days ago
fluxions/vui
View all activity

Organizations

None yet