def _init_weights(self, module):
    """ Initialize the weights """
    if isinstance(module, (nn.Linear, nn.Embedding)):
        # Slightly different from the TF version which uses truncated_normal for initialization
        # cf https://github.com/pytorch/pytorch/pull/5617
        module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
    elif isinstance(module, BertLayerNorm):
        module.bias.data.zero_()
        module.weight.data.fill_(1.0)
    if isinstance(module, nn.Linear) and module.bias is not None:
        module.bias.data.zero_()

def _init_weights(self, module):
        """ Initialize the weights.
        """
        if isinstance(module, (nn.Linear, nn.Embedding, Conv1D)):
            # Slightly different from the TF version which uses truncated_normal for initialization
            # cf https://github.com/pytorch/pytorch/pull/5617
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if isinstance(module, (nn.Linear, Conv1D)) and module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.LayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)

def _init_weights(self, module):
        """ Initialize the weights.
        """
        if isinstance(module, (nn.Linear, nn.Embedding)):
            # Slightly different from the TF version which uses truncated_normal for initialization
            # cf https://github.com/pytorch/pytorch/pull/5617
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if isinstance(module, nn.Linear) and module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, XLNetLayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)
        elif isinstance(module, XLNetRelativeAttention):
            for param in [
                module.q,
                module.k,
                module.v,
                module.o,
                module.r,
                module.r_r_bias,
                module.r_s_bias,
                module.r_w_bias,
                module.seg_embed,
            ]:
                param.data.normal_(mean=0.0, std=self.config.initializer_range)
        elif isinstance(module, XLNetModel):
            module.mask_emb.data.normal_(mean=0.0, std=self.config.initializer_range)