GitHub

MixQ_Tensorrt_LLM /quantkernel /cutlass

Stream-K, which is a new general way to do split-K. It can not only improve performance, but can also significantly reduce the number of tile sizes that need to be profiled to find the best one. Fused ...
IEEE

A Configurable FPGA Accelerator for 2D Convolution with Reduced Memory Footprint using Patch-Based General Matrix-Matrix Multiplication

Abstract: Convolution is considered as the core operation in any Convolutional Neural Network (CNN) model. However, these operations are computationally intensive and have higher latency. This problem ...
GitHub

CHANGELOG.md

Stream-K, which is a new general way to do split-K. It can not only improve performance, but can also significantly reduce the number of tile sizes that need to be profiled to find the best one. Fused ...
IEEE

2D Spectral Transposed Convolutional Neural Network for S-Parameter Predictions

Abstract: In packaging problems, S-parameter predictions are necessary. Machine learning methods lead to dimensionality related challenges which we address here through spectral trans-posed ...