TorchSparse++ - Efficient Training and Inference Framework for Sparse Convolution on GPUs
Explore a conference talk from MICRO 2023 presenting "TorchSparse++: Efficient Training and Inference Framework for Sparse Convolution on GPUs." Delve into the research conducted by Haotian Tang, Shang Yang, Zhijian Liu, and colleagues from MIT HAN Lab. Learn about their innovative approach to improving sparse convolution efficiency on GPUs for both training and inference. Discover the key features and benefits of the TorchSparse++ framework, designed to enhance performance in various applications. Gain insights into the potential impact of this technology on deep learning and computer vision tasks. Access additional resources, including the TorchSparse website, project details, and open-source code, to further explore this cutting-edge development in sparse convolution optimization.