【C++基础】运算符详解

xss666: 很有价值

数据知道: 嗯嗯 谢谢，学到了

睢志鹏: # jax.numpy与numpy的计算速度 print("jax.numpy:") x = random.uniform(random.PRNGKey(0), [10000, 10000]) start_time = time.time() res = jnp.dot(x, x) end_time = time.time() print(end_time-start_time) start_time = time.time() res = jnp.dot(x, x) end_time = time.time() print(end_time-start_time) start_time = time.time() res = jnp.dot(x, x) end_time = time.time() print(end_time-start_time) print("numpy:") x = np.random.random([10000, 10000]).astype(np.float32) start_time = time.time() res = np.dot(x, x) end_time = time.time() print(end_time-start_time) start_time = time.time() res = np.dot(x, x) end_time = time.time() print(end_time-start_time) start_time = time.time() res = np.dot(x, x) end_time = time.time() print(end_time-start_time)

睢志鹏: jax.numpy确实会比numpy快很多，你这里的矩阵维度不同，一个5000，一个10000，应该在相同维度进行比较。其次就是，jax.numpy的jit功能可以使得一次编译后速度大大提高，而numpy每次运行的速度都基本一样，具体可以试一下我的代码

1729_H: 我的理解是方便理解数学运算中的先后顺序和级别