Qua2SeDiMo learns layer-level quantization sensitivity with a GNN and constructs sub-4-bit mixed-precision weights for several diffusion models with FID close to or better than full precision.
Rooted at a dummy ‘Add’ node
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Qua$^2$SeDiMo: Quantifiable Quantization Sensitivity of Diffusion Models
Qua2SeDiMo learns layer-level quantization sensitivity with a GNN and constructs sub-4-bit mixed-precision weights for several diffusion models with FID close to or better than full precision.