vorenthos-vision-thinking-qwq-7b

Vision + Thinking Model

Fine-tuned unsloth/Qwen2.5-VL-7B-Instruct-bnb-4bit with Unsloth + QLoRA.

QwQ-VL is Qwen's dedicated reasoning vision-language model. It produces <think>…</think> chain-of-thought blocks before giving its final answer — ideal for complex visual reasoning tasks.

Training details

Usage

from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info
import torch
 
model_id = "vorenthiclabs/vorenthos-vision-thinking-qwq-7b"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    model_id, torch_dtype=torch.float16, device_map="auto"
)
 
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "image": "https://your-image-url.jpg"},
            {"type": "text",  "text": "Describe what you see and reason step by step."}
        ]
    }
]
text   = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = processor(text=[text], images=[...], return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=512)
print(processor.decode(output[0], skip_special_tokens=True))