Image-Text-to-Text
Transformers
Safetensors
qwen3_5
ocr
document-parsing
multimodal
markdown
tables
formulas
vllm
conversational
Eval Results
Instructions to use ATH-MaaS/OvisOCR2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ATH-MaaS/OvisOCR2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="ATH-MaaS/OvisOCR2") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("ATH-MaaS/OvisOCR2") model = AutoModelForMultimodalLM.from_pretrained("ATH-MaaS/OvisOCR2", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use ATH-MaaS/OvisOCR2 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "ATH-MaaS/OvisOCR2" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "ATH-MaaS/OvisOCR2", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/ATH-MaaS/OvisOCR2
- SGLang
How to use ATH-MaaS/OvisOCR2 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "ATH-MaaS/OvisOCR2" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "ATH-MaaS/OvisOCR2", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "ATH-MaaS/OvisOCR2" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "ATH-MaaS/OvisOCR2", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use ATH-MaaS/OvisOCR2 with Docker Model Runner:
docker model run hf.co/ATH-MaaS/OvisOCR2
windows上面能够正常使用OvisOCR2吗
#5
by Yunfei1994 - opened
- OvisOCR2 基于 Qwen3.5-0.8B ,它使用了 Gated Delta Network (GDN) 线性注意力层
- 这种线性注意力需要 flash-linear-attention (FLA) 库加速——这不是普通的 flash-attn,是专门针对线性注意力的加速库
- 没有 FLA 时,回退到纯 PyTorch 实现 → 每页6分钟 (实测)
- FLA 的依赖链在 Windows 上断裂:
- 需要 PyTorch ≥ 2.7.0 → Windows cu124 最高只有 2.6.0
- 需要 Triton ≥ 3.3 → Windows 上 Triton 仍是实验性支持
- 还需要 causal-conv1d → 需要 CUDA 编译,Windows 无官方 wheel
总之: Qwen3.5 的线性注意力架构在 Windows 上跑不快 ,这是架构级别的限制,不是简单装个包能解决的。