#4747th of 49 in Open models
chinese-clip
Chinese-language CLIP models for image-text embedding and cross-modal retrieval
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 6k
- License
- MIT
- Last commit
- Mar 2026
- Language
- Jupyter Notebook
5,803 on 3 Apr 20266,019 on 23 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Chinese-CLIP provides CLIP models trained on about 200M Chinese image-text pairs, in five sizes from RN50 (77M parameters) to ViT-H/14 (958M). It offers a Python API for image and text features and similarity, plus scripts for finetuning, feature extraction, KNN retrieval and zero-shot classification. Code is built on open_clip.
Who it is for: ML engineers building Chinese image-text search or zero-shot classification
Strengths
- Five model sizes, RN50 to ViT-H/14, on Hugging Face and ModelScope
- Reports zero-shot and finetuned results on MUGE, Flickr30K-CN and COCO-CN
- Finetune options include FlashAttention, gradient accumulation, FLIP masking and gradient checkpointing
- ONNX, TensorRT and Core ML conversion scripts for deployment
Weaknesses
- Datasets must be converted to base64 TSV, JSONL and LMDB before training
- Feature extraction script supports a single GPU only
- No packaged retrieval server; the demo relies on external clip-retrieval
- Stated requirements list CUDA 10.2 or later; no Docker image
What it needs
- GPU optional
- Needs PyTorch >= 1.8.0, torchvision >= 0.9.0, modelscope (optional, for model download)
- Models: ViT-B/16, ViT-L/14, ViT-L/14-336, ViT-H/14, RN50
Quick answers
Is chinese-clip open source?
Yes. Its code is on GitHub under the MIT license.
Does chinese-clip need a GPU?
No. It runs on a CPU and can use a GPU when you have one.
Also in Open models
See all 49| Rank | Project | Score |
|---|---|---|
| 1 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 87 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 86 out of 100 |
| 3 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 77 out of 100 |
| #4 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 76 out of 100 |
| #5 | speechPyTorch toolkit for training and deploying ASR, TTS and speech LLM models | 74 out of 100 |