---
title: visionpangu
canonical_url: "https://www.modelscope.cn/models/asdfgh007/visionpangu"
md_url: "https://www.modelscope.cn/models/asdfgh007/visionpangu.md"
repository: asdfgh007/visionpangu
last_updated: 2026-03-05
pipeline_tag: visual-question-answering
tasks:
  - visual-question-answering
model_type:
  - internvl
architectures:
  - InternVLForConditionalGeneration
base_model:
  - OpenGVLab/InternVL3_5-1B
  - FreedomIntelligence/openPangu-Embedded-1B-V1.1
parameters: 1.5B
tensor_type:
  - F32
library_name:
  - safetensors
  - pytorch
frameworks:
  - Pytorch
inference_backends:
  - "deploy_task text"
  - "lmdeploy 0.9.1"
  - "lmdeploy_turbomind 0.9.1"
downloads: 22
stars: 1
---

# visionpangu

> visionpangu - asdfgh007 在 ModelScope 开源的模型。pip install transformers==4.53.3 torch==2.9.1 torchvision==0.24.1 Pillow sentencepiece

asdfgh007/visionpangu 是 ModelScope 魔搭社区上的 1.5B 参数visual-question-answering模型，基于 OpenGVLab/InternVL3_5-1B、FreedomIntelligence/openPangu-Embedded-1B-V1.1 构建，可用 deploy_task text、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。

- **Repository**: asdfgh007/visionpangu
- **Tasks**: visual-question-answering
- **Parameters**: 1.5B
- **Base model**: OpenGVLab/InternVL3_5-1B, FreedomIntelligence/openPangu-Embedded-1B-V1.1
- **Inference backends**: deploy_task text, lmdeploy 0.9.1, lmdeploy_turbomind 0.9.1
- **Downloads**: 22
- **Stars**: 1
- **Last updated**: 2026-03-05

Source: https://www.modelscope.cn/models/asdfgh007/visionpangu

---

python ==3.10

pip install transformers==4.53.3 torch==2.9.1 torchvision==0.24.1 Pillow sentencepiece

运行示例：
```
import torch
from PIL import Image
from transformers import AutoProcessor,AutoModelForCausalLM, AutoTokenizer,GenerationConfig

model_path="your model save path"
image_path="img/example.jpg"

model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.float16, trust_remote_code=True).cuda()
model = model.float()

processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True)

tokenizer = AutoTokenizer.from_pretrained(
    model_path, 
    use_fast=False, 
    trust_remote_code=True,
    local_files_only=True
    )

prompt = 'Describe this image in detail.'
image= Image.open(image_path).convert("RGB")
sys_prompt = "you are visionpangu,a multimodal assistant."
messages = [
    {"role": "system", "content": sys_prompt}, # define your system prompt here
    {"role": "user", "content": [{"type":"image", "image": image},{"type":"text", "text": prompt}]}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
inputs = processor(image, text, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=1024)
result = processor.decode(output[0]).split("[unused9]助手：")[-1].strip("[unused10]")
# Preparation for inference
print(result)
```