Error: Divided by 0.

[woriazzc]

Thanks for your the implementation!

In LsqQuan.

lsq-net/quan/quantizer/lsq.py

Line 54 in 2c24a96

x = x / s_scale

If s_scale contains 0, there will be Divided by 0 Error, and the training loss will become nan.
Maybe eps should be used.

[zhutmost]

Looks that it is very rarely seen. s_scale is a floating pointed number, it generally does not become 0 during training. I will try to add an eps, when I have spare time. (it need some experiments)

[lippman1125]

Looks that it is very rarely seen. s_scale is a floating pointed number, it generally does not become 0 during training. I will try to add an eps, when I have spare time. (it need some experiments)

Not rarely.

[flymmmfly]

Not rarely.