• attention — The mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.

Latent diffusion is why high-resolution image generation became trainable outside big labs: compress away imperceptible detail first, run the diffusion model in latent space, and condition with cross-attention on text or boxes.

• articleNo Idle Gpus Managing Research Compute At Runway
• article60x Faster Cold Starts Treating Peer Gpus As Weight Servers
• articleDtensor Distributed Training
• articleStop Writing Yaml Configuring Ml Systems With Confingy

• clipLatent diffusion: compress with a VAE, diffuse in latent spaceJulia Turc
• articleWhat are Diffusion Models?Lilian Weng
• clipLatent-space text diffusion still needs an autoregressive decoderJulia Turc

Checking sign-in…

Loading comments…