A data-free method that tunes only the diagonal scales of Winograd transform matrices enables accurate 8-bit fully quantized Winograd convolution for diffusion models and ResNets.
Towards efficient and accurate winograd convolution via full quantization
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Data-Free Group-Wise Fully Quantized Winograd Convolution via Learnable Scales
A data-free method that tunes only the diagonal scales of Winograd transform matrices enables accurate 8-bit fully quantized Winograd convolution for diffusion models and ResNets.