That said, when PyTorch is instructed to free a GPU tensor it tends to cache that GPU memory for a while ... 在报错哪一行前面加上下面两行:释放无关的内存.
確定! 回上一頁