Data

train data is similarity sentence data from E-commerce dialogue, about 50w sentence pairs.

Model

model created by sentence-tansformers，model struct is bi-encoder

Usage

>>> from sentence_transformers import SentenceTransformer, util
>>> model = SentenceTransformer("tuhailong/bi_encoder_roberta-wwm-ext", device="cuda:1") 
>>> model.max_seq_length=32
>>> sentences = ["今天天气不错", "今天心情不错"]
>>> embeddings1 = model.encode([sentences[0]], convert_to_tensor=True)
>>> embeddings2 = model.encode([sentences[1]], convert_to_tensor=True)
>>> scores = util.cos_sim(embeddings1, embeddings2).cpu().numpy()
>>> print(scores)

Code

train code from https://github.com/TTurn/bi-encoder

PS

Because add the pooling layer and dense layer after model，has folders in model files. So here will be additional files "1_Pooling-config.json", "2_Dense-config.json" and "2_Dense-pytorch_model.bin". after download these files, rename them as "1_Pooling/config.json", "2_Dense/config.json" and "2_Dense/pytorch_model.bin".