aimet
aimet copied to clipboard
What is the paper referenced by QAT with Range Learning?
@666DZY666 the references are LSQ (https://arxiv.org/abs/1902.08153) and LSQ+ (https://arxiv.org/abs/2004.09576 but AIMET optimizes min/max instead of scale offset so our implementation is not exactly the same.
@666DZY666 the references are LSQ (https://arxiv.org/abs/1902.08153) and LSQ+ (https://arxiv.org/abs/2004.09576 but AIMET optimizes min/max instead of scale offset so our implementation is not exactly the same.
Why does aimet optimize min/max instead of scale/offset, is this better?
@666DZY666 the references are LSQ (https://arxiv.org/abs/1902.08153) and LSQ+ (https://arxiv.org/abs/2004.09576 but AIMET optimizes min/max instead of scale offset so our implementation is not exactly the same.
Why does aimet optimize min/max instead of scale/offset, is this better?
There are obviously some mathematical differences. However, I believe it does not matter. So its an implementation choice.