It looks like by pruning by a factor of 0.5, you reduce the size of the model by 50%? In practice, what is the expected observed change in the output before and after pruning?
The expected and observed change is virtually none. That's the whole point!
Notably, quantizing weights from 16bit weights to 4bit weights (reducing the size by 75%) also has almost no change in output quality when using modern algorithms like GPTQ.
A 0.01% loss in quality for a 4x speed up and 4x less VRAM/RAM requirement.
90GB models now fit and run on a $600 consumer video card with quality so similar the difference is only detectable on hours long automated tests with tens of thousands of iterations.