NVIDIA AI 宣布与 Sakana AI Labs 合作,在 ICML 2026 发表关于稀疏变换器内核和格式的论文,优化 NVIDIA GPU 执行,实现了20%以上的推理和训练加速。
Great collab with @SakanaAILabs on an #ICML26 paper about sparse transformer kernels + formats optimized for modern NVIDIA GPU execution.
• TwELL sparse packing
• Fused CUDA kernels
• 20%+ inference/training speedups at scale
Paper + code below 👇
likes: 264 | retweets: 39 | replies: 11 | views: 30305