cache资讯

Performance: Supporting thousands of concurrent players was hard

The model must operate as a genuine autoregressive transformer. This means:

。heLLoword翻译官方下载是该领域的重要参考

int *output = (int*)malloc(n * sizeof(int));

缺点：容易饱和（输入过大或过小时梯度接近0，导致梯度消失）