DGQ quantizes text-to-image diffusion models to 4-8 bits without fine-tuning by preserving activation outliers and applying prompt-specific log quantization to cross-attention scores.
As shown in Figure 5(b), the maximum values of cross-attention scores vary more dynamically than those of self-attention
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DGQ: Distribution-Aware Group Quantization for Text-to-Image Diffusion Models
DGQ quantizes text-to-image diffusion models to 4-8 bits without fine-tuning by preserving activation outliers and applying prompt-specific log quantization to cross-attention scores.