1344dee35dbacaaaaabdaf452f0dfe74e3ab50e4,torchsample/modules/constraints.py,MaxNorm,__call__,#MaxNorm#Any#,42

Before Change

    def __call__(self, p):
        norms = K.sqrt(K.sum(K.square(p), axis=self.axis, keepdims=True))
        desired = K.clip(norms, 0, self.m)
        p *= (desired / (K.epsilon() + norms))
        return p

    def get_config(self):
        return {"name": self.__class__.__name__,
                "m": self.m,

After Change

        self.module_filter = module_filter

    def __call__(self, module):
        if hasattr(module, "weight"):
            w = module.weight.data
            norm = torch.norm(w,2,self.axis).expand_as(w) / self.m
            norm[norm<1.] = 1.
            w.div_(norm)


class NonNeg(Constraint):
    
    Constrains the weights to be non-negative.
    

Instances