NFE
A metric measuring the inference cost of diffusion models, equal to the number of denoising steps multiplied by the number of network forward passes per step, used to compare generation quality under matched computational budgets.
Timeline (last 90 days)
Under a matched NFE budget, recurrent UDLM's generation perplexity is up to 9.8% lower than that of the non-recurrent version.
NFE (Number of Function Evaluations) refers to the number of times the model needs to call a network forward computation when generating one image (or one video, or one 3D asset).
All Facts (2)
Under a matched NFE budget, recurrent UDLM's generation perplexity is up to 9.8% lower than that of the non-recurrent version.
50%UnverifiedNFE (Number of Function Evaluations) refers to the number of times the model needs to call a network forward computation when generating one image (or one video, or one 3D asset).
50%