SKEP模型汇总

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

Pretrained Weight	Language	Details of the model
skep_ernie_1.0_large_ch	Chinese	24-layer, 1024-hidden, 16-heads, 336M parameters. Trained using the Erine model ernie_1.0
skep_ernie_2.0_large_en	English	24-layer, 1024-hidden, 16-heads, 336M parameters. Trained using the Erine model ernie_2.0_large_en
skep_roberta_large_en	English	24-layer, 1024-hidden, 16-heads, 355M parameters. Trained using the RoBERTa model roberta_large_en