3 ms·
AFAIK for over-parameterized models, performing quantization or any other form of compression won't reduce accuracy by much (don't quote me on this though).
by fzliu 3y ago
AFAIK for over-parameterized models, performing quantization or any other form of compression won't reduce accuracy by much (don't quote me on this though).