towhee.models.mdmmt.bert_mmtΒΆ
Logic for the Transformer architecture used for MMT.
Classes
Self-attention layer. |
|
Construct the embeddings from word, position and token_type embeddings. |
|
Complete Bert Model (Transformer encoder). |
|
Fully-connected layer, part 1. |
|
Complete Bert layer. |
|
Bert Model. |
|
Fully-connected layer, part 2. |
|
Extraction of a single output embedding. |
|
Self-attention mechanism. |
|
Self-attention output. |