https://yestino.com/zh-CN/entities/cuda-rng-preallocate-9ffa4f与 CUDA RNG Preallocate 相关的全部事件zh-CNhttps://yestino.com/zh-CN/events/trunk-9beae8fa8cb87e424fefc7a43ddf96418286e0d4-cuda-rng-prea-ddeff3https://yestino.com/zh-CN/events/trunk-9beae8fa8cb87e424fefc7a43ddf96418286e0d4-cuda-rng-prea-ddeff3Thu, 15 Oct 2026 01:10:01 GMTLazy CUDA RNG initialization allocates seed and offset tensors on the default stream while capture is active. 来源：PyTorch Releases