Instructions to use nowdoor/Qwen3.8-27B-Inspect-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use nowdoor/Qwen3.8-27B-Inspect-v2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="nowdoor/Qwen3.8-27B-Inspect-v2") 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)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("nowdoor/Qwen3.8-27B-Inspect-v2") model = AutoModelForMultimodalLM.from_pretrained("nowdoor/Qwen3.8-27B-Inspect-v2", 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=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use nowdoor/Qwen3.8-27B-Inspect-v2 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "nowdoor/Qwen3.8-27B-Inspect-v2" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "nowdoor/Qwen3.8-27B-Inspect-v2", "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/nowdoor/Qwen3.8-27B-Inspect-v2
- SGLang
How to use nowdoor/Qwen3.8-27B-Inspect-v2 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 "nowdoor/Qwen3.8-27B-Inspect-v2" \ --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": "nowdoor/Qwen3.8-27B-Inspect-v2", "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 "nowdoor/Qwen3.8-27B-Inspect-v2" \ --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": "nowdoor/Qwen3.8-27B-Inspect-v2", "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 nowdoor/Qwen3.8-27B-Inspect-v2 with Docker Model Runner:
docker model run hf.co/nowdoor/Qwen3.8-27B-Inspect-v2
Qwen3.8-27B-Inspect-v2
ํ๊ตญ์ด ์์ค๋ฌผ ์์ ์ ๊ฒยท์ ๋ฐ์์ ์ง๋จ ๋ณด๊ณ ์์ ์๋ฏธ ๊ฒํ ๋ฅผ ์ํ ์คํ์ฉ ๋ชจ๋ธ์
๋๋ค. ๊ณต๊ฐ v1(Qwen3.8-27B-Inspect-v01)์ ํ์ ํ์ต ๋ชจ๋ธ์ด๋ฉฐ, ๊ธฐ๋ณธ ์ถ๋ก ์ค์ ์ enable_thinking=True, reasoning_effort="medium"์
๋๋ค.
Key results
์์ค๋ฌผ ํ๊ฐ๋ ํ์ต ์๋ฃ์ ์ ํ ์ค๋ณต์ด ์๋ ํฉ์ฑ ์ฌ๋ก 50๊ฑด์ thinking-on, greedy decoding์ผ๋ก ์คํํ์ต๋๋ค. AIME๋ ๋๋ฉ์ธ ํ์ธํ๋ ํ ์ผ๋ฐ ์ถ๋ก ์ฑ๋ฅ์ด ์ ์ง๋๋์ง ํ์ธํ๊ธฐ ์ํ ์งํ์ ๋๋ค.
| ํ๊ฐ | v1 | v2 |
|---|---|---|
| ์์ค๋ฌผ 6๋จ๊ณ ํ์ ์ ํ๋ | 30/50 (60%) | 35/50 (70%) |
| ์์ค๋ฌผ macro F1 | 0.463 | 0.531 |
์ ์ </think> ํ๋กํ ์ฝ |
0/50 | 50/50 |
| ์๊ฒฉ JSON ์คํค๋ง | 46/50 | 50/50 |
| ํ์ ํฌํจ ์์ ์ด์ ์ฑ๊ณต | 25/50 | 35/50 |
| AIME 2024+2025 ์ ํ๋ | 46/60 (76.7%) | 49/60 (81.7%) |
| AIME ์ถ๋ ฅ ํ ํฐ ํ๋ ๋๋ฌ | 25/60 | 17/60 |
v2์ ์์ค๋ฌผ ํ์ ์ ํ๋๋ v1๋ณด๋ค 10.0%p ๋์๊ณ , 50๊ฑด ๋ชจ๋ ์ ์์ ์ธ thinking ์ข ๋ฃ์ ์๊ฒฉ JSON ์คํค๋ง๋ฅผ ์ถฉ์กฑํ์ต๋๋ค.
AIME ์ ํ๋๋ ์๋ณธ Base 47/60 (78.3%), v1 46/60 (76.7%), v2 49/60 (81.7%)์ด์์ต๋๋ค. ์ด๋ฒ ํ๊ฐ์์๋ ์์ค๋ฌผ ๋๋ฉ์ธ ํ์ธํ๋ ์ดํ์๋ ์ผ๋ฐ ์ถ๋ก ์ฑ๋ฅ์ด ํฌ๊ฒ ์ ํ๋์ง ์์์ต๋๋ค. ๋ค๋ง 60๋ฌธํญ์ ๋จ์ผ greedy pass ๊ฒฐ๊ณผ์ด๋ฏ๋ก ์ฑ๋ฅ ํฅ์๋ณด๋ค๋ ์ฑ๋ฅ ๋ณด์กด ์งํ๋ก ํด์ํด์ผ ํฉ๋๋ค.
Usage
Qwen3.5 ์ํคํ ์ฒ๋ฅผ ์ง์ํ๋ Transformers ๋ฒ์ ์ ์ฌ์ฉํ์ญ์์ค. ์ด ์ฒดํฌํฌ์ธํธ๋ Transformers 5.14.1/5.15.0๊ณผ vLLM 0.17.1์์ ๊ฒ์ฆํ์ต๋๋ค.
import torch
from transformers import AutoModelForImageTextToText, AutoProcessor
model_id = "nowdoor/Qwen3.8-27B-Inspect-v2"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForImageTextToText.from_pretrained(
model_id,
dtype=torch.bfloat16,
device_map="auto",
)
messages = [
{
"role": "system",
"content": (
"๋๋ ์์ค๋ฌผ ์์ ์ ๊ฒยท์ ๋ฐ์์ ์ง๋จ ๋ณด๊ณ ์๋ฅผ ์ง์นจ์ ๋ฐ๋ผ "
"์๋ฏธ ์ค์ฌ์ผ๋ก ๊ฒํ ํ๋ ์ ๋ฌธ๊ฐ๋ค. ์
๋ ฅ์ ์๋ ์ฌ์ค์ ์ถ์ ํ์ง ์๋๋ค."
),
},
{
"role": "user",
"content": "๋ณด๊ณ ์ ๋ฐ์ท์ ๊ฒํ ๊ธฐ์ค์ ์ฌ๊ธฐ์ ์
๋ ฅํฉ๋๋ค.",
},
]
prompt = processor.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=True,
reasoning_effort="medium",
)
inputs = processor(text=prompt, return_tensors="pt").to(model.device)
with torch.inference_mode():
output = model.generate(**inputs, max_new_tokens=4096, do_sample=False)
new_tokens = output[:, inputs["input_ids"].shape[1]:]
print(processor.batch_decode(new_tokens, skip_special_tokens=True)[0])
BF16 ๊ฐ์ค์น๋ ์ฝ 55.6GB์ ๋๋ค.
Intended use
- ์์ค๋ฌผ ๋ณด๊ณ ์์ ์ง์นจ ์ค์ ์ฌ๋ถ์ ๋ํ ๋ณด์กฐ ๊ฒํ
- ๊ทผ๊ฑฐ ์กฐํญ๊ณผ ๋ณด๊ณ ์ ์ธ์ฉ์ ํฌํจํ ๊ตฌ์กฐํ JSON ์์ฑ
- ๊ฒํ ์ด์ ์์ฑ๊ณผ ํ์ง๊ด๋ฆฌ ์ง์
์์ค๋ฌผ ์์ ๋ฑ๊ธ, ๋ฒ์ ์ ํฉ์ฑ ๋๋ ํ์ฅ ๊ธฐ์ ์์ ์ต์ข ํ๋จ์ ๋์ฒดํ์ง ์์ต๋๋ค.
Limitations
- ์์ค๋ฌผ ํ๊ฐ๋ ํฉ์ฑ ์ฌ๋ก 50๊ฑด์ด๋ฉฐ ์ค์ ์ ์ฒด ๋ณด๊ณ ์์ ๋ํ ์ฑ๋ฅ์ ๋ณด์ฅํ์ง ์์ต๋๋ค.
- ์์ค๋ฌผ ๋น๊ต ๊ฒฐ๊ณผ๋ thinking-on ์กฐ๊ฑด์๋ง ํด๋นํฉ๋๋ค.
- ์ฌ๋ก๊ฐ ์ ์ ์๊ฒฉ ํ์ ๋ฑ๊ธ์ ์ถ๊ฐ ํ๊ฐ๊ฐ ํ์ํฉ๋๋ค.
- ๋ณด๊ณ ์์ ์๋ ์ฌ์ค์ด๋ ๊ทผ๊ฑฐ๋ฅผ ์์ฑํ ์ ์์ผ๋ฏ๋ก ์๋ฌธ ๋์กฐ๊ฐ ํ์ํฉ๋๋ค.
- ์ด๋ฏธ์งยท์์ ์ ๋ ฅ์ ๋ํ ๋ณ๋ ๋๋ฉ์ธ ํ๊ฐ๋ ์ํํ์ง ์์์ต๋๋ค.
- AIME ๊ฒฐ๊ณผ๋ 60๋ฌธํญ์ ๋จ์ผ ์คํ์ด๋ฏ๋ก ์์ ์ ์ ์ฐจ์ด์ ๋ฏผ๊ฐํฉ๋๋ค.
ํ์ตยทํ๊ฐ ์๋ฌธ๊ณผ ์คํ ๋ก๊ทธ๋ ์ด ์ ์ฅ์์์ ๋ฐฐํฌํ์ง ์์ต๋๋ค.
License
Apache License 2.0. ์๋ณธ ๋ชจ๋ธ์ ๋ผ์ด์ ์ค์ ์ฌ์ฉ ์กฐ๊ฑด๋ ํจ๊ป ํ์ธํ์ญ์์ค.
- Downloads last month
- 15
Model tree for nowdoor/Qwen3.8-27B-Inspect-v2
Base model
Qwen/Qwen3.8-27BEvaluation results
- Six-level judgment accuracy on AutoCheck independent synthetic testtest set self-reported0.700
- General reasoning retention (AIME extracted-answer accuracy) on AIME 2024 and 2025test set self-reported0.817
