Question about Small and medium model for graph classification

[Dbgsaoge]

Hi, I am interested in your work. But I have a question about your medium_model.py; it seems that in your SpecformerMedium class, you didn't apply
```
mha_eig = self.mha_norm(eig)
mha_eig, attn = self.mha(mha_eig, mha_eig, mha_eig, key_padding_mask=e_mask)
eig = eig + self.mha_dropout(mha_eig)

    ffn_eig = self.ffn_norm(eig)
    ffn_eig = self.ffn(ffn_eig)
    eig = eig + self.ffn_dropout(ffn_eig)
     ```

the encoding process. Instead, you directly apply decoder to the EE representation. Is it the right process?