TL;DR
LiteBDS-ViT, a hardware-friendly structural optimization framework that replaces dense attention and feed-forward projections with block-diagonal shift projections, allocates group numbers layer-wise under a global operation budget, and implements fused CUDA kernels to translate theoretical sparsity into practical GPU speedup is proposed.
Vision Transformers (ViTs) achieve strong performance on image recognition and mobile vision applications, but their high-dimensional linear projections and attention computations still impose substantial storage and inference costs. Extremely low-bit quantization is a promising solution, yet ViTs often suffer severe a...
Four-bit floating-point (FP4) Tensor Cores accelerate matrix multiplication, but scale computation, operand packing, layout construction, and saved backward state can erase the gain. We present \emph{format-aware fusion}, which co-designs each quantization producer with its scale domain and consumer layout for native \...
This work builds a GPU library that emulates SC matrix multiplication at scale, exposes stream lengths as first-class kernel arguments, and evaluates SC end-to-end on image classification, object detection and instance segmentation, class-conditional image generation, and visual world-model planning.
FluxBin is proposed, an algorithm-kernel co-design that synergizes post-training quantization with a highly optimized CUDA kernel and introduces Decoupled Row-Column Binary Decomposition to enhance representational capacity while maintaining hardware efficiency.
Deploying vision-transformer backbones on edge devices requires low-bit quantization, and mixed-precision methods must assign each layer a bit-width under a memory budget. Allocation is usually driven by per-layer sensitivity estimated from loss curvature (Hessian traces), following HAWQ. We propose CASQ (Configuration...
This work identifies and formalizes the principle that organizes the Great Inversion, the Great Inversion: allocation-flexible coding rewards energy concentration, whereas the grouped shared-scale quantization a deployed matrix instruction performs rewards within-group flattening.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
