DebertaV2模型汇总#
下表汇总介绍了目前PaddleNLP支持的DebertaV2模型对应预训练权重。
Pretrained Weight |
Language |
Details of the model |
---|---|---|
|
English |
24-layer, 1024-hidden, 16-heads, 304M parameters. The deberta-v3-large model fine-tuned using the SQuAD2.0 dataset. |
|
English |
24-layer, 1536-hidden, 24-heads, 900M parameters. The deberta-v2 model. |
|
English |
12-layer, 768-hidden, 12-heads, 86M parameters. The deberta-v2 model. |
|
English |
24-layer, 1024-hidden, 16-heads, 304M parameters. The deberta-v2 model. |