Restoring Diversity in Distilled Image and Video Generators by Probing Activations

Research Hugging Face Papers Source language: Chinese

论文摘要(作者报告,未经本站复测):Distilling diffusion models into few-step student models substantially reduces sampling latency, but induces severe diversity collapse: samples generated from different random seeds show minimal…

Excerpt

论文摘要(作者报告,未经本站复测):Distilling diffusion models into few-step student models substantially reduces sampling latency, but induces severe diversity collapse: samples generated from different random seeds show minimal…

Short excerpt from the source — open the original for the full text.

Sources

← Back to the daily brief