I get on Exercise 3 this error for unitests :
Failed test case: Error count greater than 80%.
Expected: Error count < 80%
Got: Error count = 40.00
It doesn’t make sense 
Did anyone passed those unitests ?
Hi!
You’re right, that doesn’t make sense. The text should say “accuracy less than 80%” instead of “error count greater than 80%”. Also, there’s a mistake in the calculations. In your case, you got 2 incorrect results, but you should get at most 1 incorrect result. I’ll fix the unittest to address this.
In the meantime, you might want to tweak your query to improve the results. Here’s a suggestion, check out the example right below the unittests, it runs the same sample as the test itself. Then you can see which question was labeled incorrectly and try improving your prompt by adding a similar question.
I’ll let you know once the unittest is fixed, but this shouldn’t prevent you from getting proper grading upon submission.
Thanks,
Lucas
Hi @lucas.coutinho , I hope you are well. I’m just following up on this issue since it appears I’m struggling with potentially the same issue. Also, I’ve tried multiple shorter versions of the prompt for exercise 3 but it seems 150 tokens is too small. I just wanted to check 150 was the intended value. In the previous exercise the token limit was 180 and I didn’t run into any issues there.
Thanks!
Hi!
I just fixed the unittests to now show the error count directly and not a percentage. I also increased the token limit to 170.
This was a silent update so you might need to refresh your workspace to get the most updated version.
I hope that helps!
Cheers,
Lucas
Thanks @lucas.coutinho 
! You are a rock star. Appreciate your promptly response.