I passed all the tests before this exercise 8, but I can not figure out the problem here :
Notice that your cost values after 1000 and 2000 iterations are quite a bit higher than the expected values. That would suggest there is something wrong with your “update parameters” logic. But if you are sure that all your previous functions passed their tests, then the bug (whatever it is) must be in nn_model . E.g. you are passing the wrong arguments to one of the previous functions when you call it from nn_model .
Did you set “learning_rate” to “0.005” at somewhere ? All your outputs are exactly same as the case if we set the learning rate to 0.005. Please double check.
Yes, I had set “learning_rate” to “0.005”.
You do not need to set by yourself for this exercise. (In the real world, you need to set as a hyper-parameter, but, for this testing, do not.) The purpose of this test is to check your implementation.
So, just removing it makes you go forward. 
Hi, Thanks a lot, it is as you said, need to remove the learning_rate setting. It works! And thanks again for pointing out the local/global variable errors !
I was wondering if you could resolve the problem.
Thanks in advance
The solution is in this thread. Please read the full thread.