If you have a numpy array and want to avoid a copy, use torch.as_tensor() . A tensor of specific data type ... Returns a copy of this object in CUDA memory.
確定! 回上一頁