Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
ohai.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
A cozy, fast and secure Mastodon server where everyone is welcome. Run by the folks at ohai.is.

Administered by:

Server stats:

1.8K
active users

ohai.social: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.4.0-nightly.2025-04-01

#diffusionmodels

4 posts3 participants0 posts today
Public
Chris Green

An attempt at a nostalgic 1960s photograph . After many iterations I had an image which had the structure I wanted but the finer details were all wrong so I used that image as a controlnet, evolving the text prompt and other settings to create the final image. This is based on Colossus Project Flux v5 with multi-stage sampling, no LoRAs and CannyEdge Pre-processing. #nostalgia #younglove #GenAI #diffusionmodels #dailydoodle #imagegeneration #comfyUI #flux #fluxai
diffusiondoodles.substack.com

Public
N-gated Hacker News

In today's episode of "How to Make #Gibberish Sound Smart," we dive into an article that bravely attempts to bridge the vast chasm between #autoregressive models and diffusion models with absolutely zero context 🧠🔍. Meanwhile, Hugging Face is here to remind us that no amount of tech jargon can hide the fact they're just as desperate for your money as your local neighborhood pyramid scheme 💸🎩.
arxiv.org/abs/2503.09573 #techjargon #diffusionmodels #HuggingFace #HackerNews #ngated

arXiv logo
arXiv.orgBlock Diffusion: Interpolating Between Autoregressive and Diffusion Language ModelsDiffusion language models offer unique benefits over autoregressive models due to their potential for parallelized generation and controllability, yet they lag in likelihood modeling and are limited to fixed-length generation. In this work, we introduce a class of block diffusion language models that interpolate between discrete denoising diffusion and autoregressive models. Block diffusion overcomes key limitations of both approaches by supporting flexible-length generation and improving inference efficiency with KV caching and parallel token sampling. We propose a recipe for building effective block diffusion models that includes an efficient training algorithm, estimators of gradient variance, and data-driven noise schedules to minimize the variance. Block diffusion sets a new state-of-the-art performance among diffusion models on language modeling benchmarks and enables generation of arbitrary-length sequences. We provide the code, along with the model weights and blog post on the project page: https://m-arriola.com/bd3lms/
Public
Hacker News

Block Diffusion: Interpolating Between Autoregressive and Diffusion Models

arxiv.org/abs/2503.09573

arXiv logo
arXiv.orgBlock Diffusion: Interpolating Between Autoregressive and Diffusion Language ModelsDiffusion language models offer unique benefits over autoregressive models due to their potential for parallelized generation and controllability, yet they lag in likelihood modeling and are limited to fixed-length generation. In this work, we introduce a class of block diffusion language models that interpolate between discrete denoising diffusion and autoregressive models. Block diffusion overcomes key limitations of both approaches by supporting flexible-length generation and improving inference efficiency with KV caching and parallel token sampling. We propose a recipe for building effective block diffusion models that includes an efficient training algorithm, estimators of gradient variance, and data-driven noise schedules to minimize the variance. Block diffusion sets a new state-of-the-art performance among diffusion models on language modeling benchmarks and enables generation of arbitrary-length sequences. We provide the code, along with the model weights and blog post on the project page: https://m-arriola.com/bd3lms/
#HackerNews#Block#Diffusion
ExploreLive feeds
About