------------------------------------
TinyBert模型汇总
------------------------------------



下表汇总介绍了目前PaddleNLP支持的TinyBert模型以及对应预训练权重。
关于模型的具体细节可以参考对应链接。

+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+
| Pretrained Weight                                                                | Language     | Details of the model                                                             |
+==================================================================================+==============+==================================================================================+
|``tinybert-4l-312d``                                                              | English      | 4-layer, 312-hidden,                                                             |
|                                                                                  |              | 12-heads, 14.5M parameters.                                                      |
|                                                                                  |              | The TinyBert model distilled from                                                |
|                                                                                  |              | the BERT model ``bert-base-uncased``                                             |
+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+
|``tinybert-6l-768d``                                                              | English      | 6-layer, 768-hidden,                                                             |
|                                                                                  |              | 12-heads, 67M parameters.                                                        |
|                                                                                  |              | The TinyBert model distilled from                                                |
|                                                                                  |              | the BERT model ``bert-base-uncased``                                             |
+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+
|``tinybert-4l-312d-v2``                                                           | English      | 4-layer, 312-hidden,                                                             |
|                                                                                  |              | 12-heads, 14.5M parameters.                                                      |
|                                                                                  |              | The TinyBert model distilled from                                                |
|                                                                                  |              | the BERT model ``bert-base-uncased``                                             |
+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+
|``tinybert-6l-768d-v2``                                                           | English      | 6-layer, 768-hidden,                                                             |
|                                                                                  |              | 12-heads, 67M parameters.                                                        |
|                                                                                  |              | The TinyBert model distilled from                                                |
|                                                                                  |              | the BERT model ``bert-base-uncased``                                             |
+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+
|``tinybert-4l-312d-zh``                                                           | Chinese      | 4-layer, 312-hidden,                                                             |
|                                                                                  |              | 12-heads, 14.5M parameters.                                                      |
|                                                                                  |              | The TinyBert model distilled from                                                |
|                                                                                  |              | the BERT model ``bert-base-uncased``                                             |
+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+
|``tinybert-6l-768d-zh``                                                           | Chinese      | 6-layer, 768-hidden,                                                             |
|                                                                                  |              | 12-heads, 67M parameters.                                                        |
|                                                                                  |              | The TinyBert model distilled from                                                |
|                                                                                  |              | the BERT model ``bert-base-uncased``                                             |
+----------------------------------------------------------------------------------+--------------+----------------------------------------------------------------------------------+