A mixed-precision activation quantization framework that processes unimportant windows of Swin transformers at lower bit width, improving the accuracy-efficiency trade-off versus uniform quantization and activation pruning.
Sparsevit: Revisiting activation spar- sity for efficient high-resolution vision transformer
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
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
MixA-Q: Revisiting Activation Sparsity for Vision Transformers from a Mixed-Precision Quantization Perspective
A mixed-precision activation quantization framework that processes unimportant windows of Swin transformers at lower bit width, improving the accuracy-efficiency trade-off versus uniform quantization and activation pruning.