Guo Sheng

Add ParallelExecutor for training, validation and saving in Transformer

1

Add ParallelExecutor for training, validation and saving in Transformer.

Tune the Transformer model for wmt14

1

Tune the Transformer model for wmt14

Make the Transformer network configurations more flexible

1

Make the Transformer network configurations more flexible

Refine the Transformer network to be more flexible and with higher BLEU

1

Refine the Transformer network to be more flexible and with higher BLEU

Add fluid_transformer.md

1

Add fluid_transformer.md

Fix text decoding in Transformer under python3.

1

Fix text decoding in Transformer under python3.

Rename neural_machine_translation as PaddleMT.

1

Rename neural_machine_translation as PaddleMT.

Update Transformer details

1

Update Transformer details

Add validation for dygraph Transformer.

1

Add validation for dygraph Transformer. Add cross-attention cache for dygraph Transformer. Add greedy search for dygraph Transformer. 此外若要与 T2T 一致，还请参照 #3684 中的内容进行细节上的更新。主要是position encoding、scaled_dot_product_attention中的scale