jingyaogong/clip-vit-base-patch16-ve

CLIP ViT-B/16 Vision Encoder (Chinese)

模型来源：iic/multi-modal_clip-vit-base-patch16_zh（阿里达摩院 Chinese CLIP）

仅裁剪保留了 Vision Encoder 部分，去除了 Text Encoder 及其他无关权重，并以 float16 精度保存为 HuggingFace CLIPVisionModel 兼容的 safetensors 格式。

---

Source: iic/multi-modal_clip-vit-base-patch16_zh (Alibaba DAMO Academy Chinese CLIP)

Only the Vision Encoder is retained. The Text Encoder and all unrelated weights have been removed. Weights are saved in float16 precision as HuggingFace CLIPVisionModel-compatible safetensors format.

---

Model Info

Item	Value
Architecture	ViT-B/16
Parameters	85.8M
Hidden Size	768
Layers	12
Patch Size	16
Input Resolution	224×224
Output Tokens	196 (14×14, CLS removed)
Weight Precision	float16
File Size	~164MB

Usage

from transformers import CLIPVisionModel, CLIPImageProcessor

model = CLIPVisionModel.from_pretrained("path_to/clip-base-vision-encoder")
processor = CLIPImageProcessor.from_pretrained("path_to/clip-base-vision-encoder")