Instructions to use summerAIO/Looped-Qwen-gsm8k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use summerAIO/Looped-Qwen-gsm8k with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="summerAIO/Looped-Qwen-gsm8k", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("summerAIO/Looped-Qwen-gsm8k", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use summerAIO/Looped-Qwen-gsm8k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "summerAIO/Looped-Qwen-gsm8k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "summerAIO/Looped-Qwen-gsm8k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/summerAIO/Looped-Qwen-gsm8k

SGLang

How to use summerAIO/Looped-Qwen-gsm8k 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 "summerAIO/Looped-Qwen-gsm8k" \
    --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": "summerAIO/Looped-Qwen-gsm8k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "summerAIO/Looped-Qwen-gsm8k" \
        --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": "summerAIO/Looped-Qwen-gsm8k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use summerAIO/Looped-Qwen-gsm8k with Docker Model Runner:
```
docker model run hf.co/summerAIO/Looped-Qwen-gsm8k
```

Looped-Qwen-gsm8k

summerMC/Looped-Qwen-pre (Looped Gated DeltaNet, 約 1B) を追加学習し、 LoRA を本体にマージした bf16 版です。読み込みには trust_remote_code=True が必要です。

評価 (GSM8K test 先頭 200 問, 答え部分の teacher-forcing loss)

モデル	loss	ppl
summerMC/Looped-Qwen-pre (ベース)	2.3437	10.42
このモデル	1.1382	3.12

正解率 (生成して答えを照合) ではなく、正解の答え部分に対する loss です。

使い方

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "summermc/Looped-Qwen-gsm8k"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True, dtype=torch.bfloat16).cuda().eval()
ids = tok.apply_chat_template([{"role": "user", "content": "Janet has 16 eggs. She eats 3 and bakes with 4. How many are left?"}],
                              add_generation_prompt=True, return_tensors="pt", return_dict=True,
                              enable_thinking=False)["input_ids"].cuda()
out = model.generate(ids, max_new_tokens=256, do_sample=True, temperature=0.7, top_p=0.9, repetition_penalty=1.1)
print(tok.decode(out[0, ids.shape[1]:], skip_special_tokens=True))

pip install -U transformers flash-linear-attention 推奨。greedy 生成では同じ行を繰り返しやすいので、 repetition_penalty を付けた sampling を推奨します。

注意

小型モデルのため計算ミスや繰り返しが起きます。
ライセンスは元モデル (Qwen/Qwen3.5-2B, Apache-2.0) に従います。
元モデル: Qwen Team / 全 GDN 化・Looped 化: summerMC, j-llm

Downloads last month: 350

Safetensors

Model size

1.0B params

Tensor type

BF16

Model tree for summerAIO/Looped-Qwen-gsm8k

Base model

Qwen/Qwen3.5-2B-Base

Finetuned

Qwen/Qwen3.5-2B

Finetuned

summerAIO/Looped-Qwen-pre

Finetuned

(2)

this model