General Text Embeddings (GTE) model. Towards General Text Embeddings with Multi-stage Contrastive Learning. The GTE models are trained by Alibaba DAMO Academy. They are mainly based on the BERT framework