Unable to get correct result for 2.4 - Cost Function.
Forward function and softmax results were okay but cost function results were not as expected.
Well, notice that they gave you the compute_cost code. If you look at how they constructed that test case, the other piece of your code it depends on is the forward_prop function. Maybe that is incorrect. Did you check your output values on the previous test for forward_prop?
Update: oh, sorry, I didn’t check what you posted carefully enough the first time. That’s what you showed. So it’s more of a mystery. But if all your code is correct and they gave you the compute_cost code, the only way it could be wrong is if you modified the compute_cost code. Maybe you should get a clean copy and carefully “copy/paste” over just your code and see if that fixes things.
This is a bug (the “Expected output” cell is not correct) and it will be corrected soon.
Nice job for spotting it. Thanks!
1 Like
I was able to get the same expected output as shown there.
Hmm… That is strange. Could you share your Notebook with me through DM? Maybe we have different versions.
Oh, sorry, it turns out I was stuck in an old version of the course because I enrolled a long time ago and hadn’t looked at it lately. I just switched and will have to rework everything. Stay tuned!
1 Like
Yes, I now see the same problem that Vishwam reported: the expected value does not match on the cost output. But several other problems with the earlier version of the notebook have been fixed. 
1 Like
As @arvyzukai told, it is a bug, so we can ignore it safetly.
1 Like
Also noticed this bug. Cost value in the Expected output seems to be calculated for tmp_batch_size = 1. With these value it equals 8.9542
1 Like