Muse: Text-To-Image Generation via Masked Generative Transformers · HackerPapers discussions

Muse: Text-To-Image Generation via Masked Generative Transformers — mentioned by airgapstopgap

Wed, 26 Jul 2023 20:24:01 GMT

Diffusion is more parameter-efficient and you quickly saturate the target fidelity, especially with some refiner cascade. It's a solved problem. You do not need more than maybe 4B total. Images are far more redundant than text. In fact, most interesting papers since Imagen show that you get more mileage out of scaling the text encoder part, which is, of course, a Transformer. This is what drives accuracy, text rendering, compositionality, parsing edge cases. In SD 1.5 the text encoder part (CLIP ViT-L/14) takes a measly 123M parameters.[1] In Imagen, it was T5-XXL with 4.6B [2]. I am…

Paper and discussion history. A mention is not an endorsement.

Muse: Text-To-Image Generation via Masked Generative Transformers — HN submission

Mon, 09 Jan 2023 08:28:29 GMT

Muse: Text-to-Image Generation via Masked Generative Transformers

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Muse: Text-To-Image Generation via Masked Generative Transformers — mentioned by espadrine

Sat, 07 Jan 2023 15:06:22 GMT

> The explanation in the original paper turns out not to be true; you can get rid of most of their assumptions and it still works I’ll admit it is amusing that some assumptions on why it works were incorrect. The core idea of a Markov chain[0] where each state change leads to higher likelihood, is bound to work, even if the rest doesn’t. In my mind, the Muse paper[1] gets closer to why it works: ultimately, the denoiser tries to match the latent space for an implicit encoder. The Muse system does this more directly and more effectively, by using cross-entropy loss on latent tokens instead. In…

Paper and discussion history. A mention is not an endorsement.

Muse: Text-To-Image Generation via Masked Generative Transformers — mentioned by nodja

Thu, 05 Jan 2023 13:21:49 GMT

Regardless of what the author of the repo did or did not do, I think that deleting his account and everything in it is a failure on the part of github and warrants a policy change. There's lots of good knowledge contained in the issues and discussions hosted on the webui repo all of which are now lost or very hard to find. Even papers coming from big companies have directly used the repo as reference[1]. I was personally playing around with prompt2prompt and reading issues when the ban happened and now I can no longer find the most recent comments, I still had the url but the internet archive…

Paper and discussion history. A mention is not an endorsement.

Muse: Text-To-Image Generation via Masked Generative Transformers — mentioned by minimaxir

Tue, 03 Jan 2023 19:12:52 GMT

Paper: https://arxiv.org/abs/2301.00704 This time, they didn't explicitly say that they weren't going to release it publicly.

Paper and discussion history. A mention is not an endorsement.