TODO list for the ConvS2S.

[lcy-seso]

- [ ] prepare the dataset https://arxiv.org/abs/1605.00459. a shared task with Transformer
- [ ] enhance the lookup_table operator to support the special token: padding index. #7309. a shared task with Transformer

• implement a general purpose normalization operator A general purpose normalization operator is needed. #7350. @lcy-seso
• wrap the weight normalization.
  • l2 normalize layer.
  • wrap weight normalization.
  • serialize weight normalization for inference.
• wrap the positional embedding.
• wrap the dot product attention. Add Python wrapper for dot-product-attention #7602
• wrap GLU unit. Add python wrapper for GLU #7525
• wrap deep convolution encoding and decoding block with attention.
• build the entire model.
• enhance the documentation of operators used in ConvS2S.
• add beam search for ConvS2S.
• merge the entire model into the models repo (actually can merge the work part by part).
• clip by norm