collaborators

1 paper

cs.CV2026

Timestep-Aware SVDQuant-GPTQ for W4A4 Quantization of Wan2.2-I2V

Junhao Wu, Dezhong Yao, Hai Jin

W4A4 quantization of large video diffusion Transformers offers substantial memory savings but is hindered by two main challenges: sparse large-magnitude activation outliers, and st…