tailieunhanh - Pre-training and fine-tuning ELECTRA models for various Vietnamese natural language processing tasks

In recent years, the Natural Language Processing community was impacted greatly by models based on the BERT architecture (Devlin et al., 2018). The Transformer-based Masked Language Model (MLM) has yielded significant improvement on many Natural Language Processing problems. However, it requires huge computing power and makes pre-training models a resource-consuming process. |

TỪ KHÓA LIÊN QUAN