Instructions to use TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k with Transformers:

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

pipe = pipeline("text-generation", model="TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k")
model = AutoModelForCausalLM.from_pretrained("TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k

SGLang

How to use TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k 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 "TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k" \
    --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": "TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k",
		"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 "TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k" \
        --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": "TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k with Docker Model Runner:
```
docker model run hf.co/TMLR-Group-HF/Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Majority-Voting: Llama-3.2-3B-Instruct trained on DAPO-14k

This model is a Llama-3.2-3B-Instruct checkpoint trained using the Majority-Voting method on the DAPO-14k training set. It is part of the research presented in the paper Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models.

The Co-rewarding framework introduces a novel self-supervised reinforcement learning (RL) approach designed to enhance the reasoning capabilities of Large Language Models (LLMs). It improves training stability by seeking complementary supervision from different perspectives, thereby addressing issues like training collapse and reward hacking often seen in other self-rewarding methods. The framework was empirically shown to achieve stable training and outperform other self-rewarding baselines on various mathematical reasoning benchmarks.

For more details, code, and other checkpoints, refer to the official GitHub repository: https://github.com/tmlr-group/Co-rewarding.