What I think the answer is supposed to be:
1 person lied. Shout out to @Callidus for pointing out the somewhat obvious implication of my logic
As I'll explain, there is no single question that can satisfy the requirements for the "amazing question". Hence OP lied in saying that such a question exists.
What I think the answer really is:
This question does not provide enough information for us to know how many people lied. As I show in my "alternative theory" section, there are compound questions that actually work. However, without knowing which question is used and the order in which the husband and wife are spoken to, we have no way to tell how many knights and knaves there are - you could formulate a question for which a husband saying "Yes" followed by the wife saying "No" means they are both knights while a wife saying "No" followed by the husband saying "Yes" means they are both knaves. So you could have 100 knaves, 100 knights, or any combination of knights and knaves.
ABOUT THE "AMAZING QUESTION"
There's a reason why the amazing question is not given to us:
No such real question could possibly exist unless it is asymmetrical!
Let's first tackle some properties of the question. We know that there are 4 possible pairings, KK, KN, NK, and NN (K=knight, N=knave). Based on the requirements of the question, we know that if you ask one person you get no conclusive information, but if you ask their spouse then you know both of their tribes. Let's focus on the husbands first:
Because we won't know his tribe based on his answer, any answer that he can give as a knight must be something he can give as a knave. Additionally, if his wife is of a particular tribe his response must change depending on his tribe. Because we won't know his wife's tribe based on his answer, his response must change depending on her tribe.
So if he's a knight, he'll say one thing (A) if she's a knight, and something else (B) if she's a knave. So if he's a knave, he'll say either A or B. Since he can't say A if she's a knight, he'll say B. Then, he'll say A if she's a knave.
H W | R
K K | A
K N | B
N K | B
N N | A
The same logic holds for the wife:
H W | R
K K | C
K N | D
N K | D
N N | C
Here's where we see the "amazing" part of the question:
H W | R R
K K | A C
K N | B D
N K | B D
N N | A C
In order to satisfy the requirements that asking a single person gives you no information about their two tribes, we've forced the question to be in a form such that asking the spouse gives you no additional information. Asking once allows you to know whether or not the two are from the same tribe, but then you can gather no additional information by using the question.
This riddle can't be solved by attempting to derive the "amazing question", because it cannot exist. Since the question can't exist, there is no way to solve the puzzle in its current form - why should the husbands and wives have to give symmetrical answers? Perhaps in a KK couple, the husband will answer 'Yes' while the wife answers 'No'. Or perhaps that's what a NN couple would answer. So we could have either everyone telling the truth, or no-one.
An alternative theory
There are asymmetrical questions that make this work, such as having the answer change depending on whether or not you have talked to their spouse.
A possible truth-table for the question under these conditions:
1 2 | R R
K K | Y Y
K N | N Y
N K | N N
N N | Y N
A question that fits this would be "If your spouse just answered my question did they say 'yes' otherwise if I asked you if you and your spouse are from the same tribe, would you say yes?"
1 2 | R R
K K | Y N
K N | N N
N K | N Y
N N | Y Y
A question that fits this would be "If your spouse just answered my question did they say 'no' otherwise if I asked you if you and your spouse are from the same tribe, would you say yes?"
Comparing the two possibilities:
This still doesn't help us solve the riddle. This allows the question to exist, but we have no way to determine how many people lied. If it is the second question and we always ask the husband first, then he will say "Yes" and his wife will say "No" if they are both knights (nobody lies). If we ask the wife first and then the husband, she will say "No" and he will say "Yes" if she is a knave and he is a knight (50 people lie). If we're not necessarily consistent in who we ask first, then we could have anywhere between 0 and 50 liars. It wouldn't be too hard to formulate another question so that the husband saying "Yes" and the wife saying "No" could either be two knaves or a knight and a knave, depending on who is asked first. Then we could have between 50 and 100 liars. In short, nobody or everyone could be liars depending on the order in which they are asked and on what the question is.
The reason I considered this an alternate theory instead of my main theory is because it feels like cheating - we're essentially asking two different questions. Also, if you object to the "if ... otherwise ..." construction of the questions, you can put it into a form that only uses 'and's and 'or's - "Is it true that 1. I have talked to your spouse and they said x, or 2. I have not talked to your spouse and if I asked you...". (might not be exactly the way to translate it, but I don't feel like going through it to make sure right now)