GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
vankey
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
vankey
Model Tree
Qwen/Qwen2.5-VL-7B-Instruct
Input Modalities
Output Modalities
Supported Functionality
| Base model | Qwen2.5-VL-7B-Instruct |
| Architecture | Qwen2_5_VLForConditionalGeneration |
| Inference resolution | 1344 × 896 (W × H) |
| Precision (weights) | bfloat16 |
| Precision (compute) | float32 (load with torch_dtype=torch.float32) |
| Max new tokens | 8192 |
The weights are stored in bfloat16 (~16 GB). Always load them with
torch_dtype=torch.float32so computation runs in float32 (the bf16 weights are upcast on load). The base model is not bundled here — download it from Qwen/Qwen2.5-VL-7B-Instruct if needed. This repository only releases the fine-tuned DocShield-7B weights.
Install dependencies:
bash
pip install transformers torch torchvision pillow opencv-python qwen-vl-utilsRun inference (see inference.py in this repo):
bash
python inference.py --image path/to/image.jpgpython
import cv2, torchfrom PIL import Imagefrom transformers import AutoProcessor, Qwen2_5_VLForConditionalGenerationfrom qwen_vl_utils import process_vision_info MODEL_PATH = "vankey/DocShield-7B" model = Qwen2_5_VLForConditionalGeneration.from_pretrained( MODEL_PATH, torch_dtype=torch.float32, attn_implementation="eager", device_map="auto",)model.eval()processor = AutoProcessor.from_pretrained(MODEL_PATH) SYSTEM_PROMPT = ( "你是一个图像鉴伪专家，擅长结合视觉，文字结合伪造特征分析手段鉴别输入图像的真假。" "分析过程中，你会逐步分析，抽丝剥茧，找到图像伪造的蛛丝马迹，最终给出专业的鉴别结果及分析。")USER_PROMPT = "请帮我分析这张图片是否是伪造的,并给出分析报告." image = cv2.cvtColor(cv2.resize(cv2.imread("image.jpg"), (1344, 896)), cv2.COLOR_BGR2RGB)image = Image.fromarray(image) messages = [ {"role": "system", "content": SYSTEM_PROMPT}, {"role": "user", "content": [ {"type": "image", "image": image}, {"type": "text", "text": USER_PROMPT}, ]},] text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)image_inputs, video_inputs = process_vision_info(messages)inputs = processor(text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt").to(model.device) with torch.no_grad(): out = model.generate(**inputs, do_sample=True, temperature=1.0, top_p=1.0, top_k=0, repetition_penalty=1.0, max_new_tokens=8192) generated = out[:, inputs["input_ids"].shape[1]:]print(processor.batch_decode(generated, skip_special_tokens=True)[0])These choices are required to reproduce the reported results:
1344 × 896 (W × H) before processing:
cv2.resize(image, (1344, 896)).float32, never compute in bfloat16. The weights are
stored as bfloat16; load them with torch_dtype=torch.float32 so they are upcast
and computation runs in float32. Computing in bfloat16 accumulates rounding error
over the long CoT and degrades output into gibberish on harder images.
float16 + eager attention also overflows (NaN) for long contexts.
float32 (compute) + eager is the verified configuration.temperature=1.0, top_p=1.0, top_k=0, repetition_penalty=1.0
(full multinomial sampling). Do not use the values in the bundled
generation_config.json (text
temperature=0.1, top_k=1, top_p=0.001, repetition_penalty=1.05bibtex
@article{docshield2026, title={DocShield: A Forensic Vision-Language Model for Document Forgery Analysis}, author={DocShield}, year={2026}, url={https://arxiv.org/abs/2604.02694}}Apache-2.0.
attn_implementation="eager"