Instructions to use yangxw/Llama-3.2-1B-countdown-backtrack with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yangxw/Llama-3.2-1B-countdown-backtrack with Transformers:

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

pipe = pipeline("text-generation", model="yangxw/Llama-3.2-1B-countdown-backtrack")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("yangxw/Llama-3.2-1B-countdown-backtrack")
model = AutoModelForCausalLM.from_pretrained("yangxw/Llama-3.2-1B-countdown-backtrack")

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use yangxw/Llama-3.2-1B-countdown-backtrack with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yangxw/Llama-3.2-1B-countdown-backtrack"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yangxw/Llama-3.2-1B-countdown-backtrack",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/yangxw/Llama-3.2-1B-countdown-backtrack

SGLang

How to use yangxw/Llama-3.2-1B-countdown-backtrack 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 "yangxw/Llama-3.2-1B-countdown-backtrack" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yangxw/Llama-3.2-1B-countdown-backtrack",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "yangxw/Llama-3.2-1B-countdown-backtrack" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yangxw/Llama-3.2-1B-countdown-backtrack",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use yangxw/Llama-3.2-1B-countdown-backtrack with Docker Model Runner:
```
docker model run hf.co/yangxw/Llama-3.2-1B-countdown-backtrack
```

Model Card for Model ID

Self-Backtracking: A novel self-backtracking method for improving language model reasoning, as described in Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models.

Model Details

Developed by: Xiao-Wen Yang and Xuan-Yi Zhu and Wen-Da Wei and Ding-Chu Zhang and Jie-Jing Shao and Zhi Zhou and Lan-Zhe Guo and Yu-Feng Li
Model type: Llama
Language(s) (NLP): en
License: mit
Finetuned from model: Llama 3.2

Model Sources

Repository: https://github.com/LAMDASZ-ML/Self-BackTracking
Paper: Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models

Uses

The integration of slow-thinking mechanisms into large language models (LLMs) offers a promising way toward achieving Level 2 AGI Reasoners.

How to Get Started with the Model

Use the code below to get started with the model.

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("yangxw/Llama-3.2-1B-countdown-backtrack")
model = AutoModelForCausalLM.from_pretrained("yangxw/Llama-3.2-1B-countdown-backtrack")

prompt = "What is 2 + 2?"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0]))

Downloads last month: 4

Safetensors

Model size

1B params

Tensor type

F32

Paper for yangxw/Llama-3.2-1B-countdown-backtrack

Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models

Paper • 2502.04404 • Published Feb 6, 2025 • 25