#7171st of 80 in Open models
text2vec
Python library for sentence embeddings and Chinese text similarity models
- Stars
- 5k
- License
- Apache-2.0
- Last commit
- Feb 2026
- Last release
- Sep 2023
- Language
- Python
4,923 on 22 Apr 20264,976 on 11 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
text2vec is a PyTorch-based Python library that encodes words, sentences and paragraphs into vectors. It implements Word2Vec, SBERT, CoSENT and BGE training and inference, and links pretrained Chinese and multilingual CoSENT models on Hugging Face, benchmarked with Spearman scores on ATEC, BQ, LCQMC, PAWSX and STS-B. It also supports multi-GPU/CPU inference and a CLI for batch encoding.
Who it is for: Engineers building Chinese or multilingual semantic matching pipelines
Strengths
- Pretrained Chinese and multilingual CoSENT models, with published Spearman scores on seven test sets
- Covers training and inference for SBERT, CoSENT and BGE in one package
- Multi-GPU and multi-CPU inference, plus a CLI for batch encoding
- Models also load via plain transformers or sentence-transformers, so the library is optional
Weaknesses
- Last release v1.2.9 is from September 2023; newer embedding models are not covered
- README and docs are mostly in Chinese; English version is a separate file
- Word2Vec vector file link points to Baidu Pan, awkward outside China
- Released models are 2023-era BERT-base size, except one bge-large variant
What it needs
- GPU optional
- Needs PyTorch, transformers, gensim
- Models: Word2Vec, SBERT, CoSENT, BGE, Hugging Face transformers models
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Open models
See all 80| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 81 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 78 out of 100 |