Instructions to use PIPer-iclr/PIPer-8B-RL-only with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PIPer-iclr/PIPer-8B-RL-only with Transformers:

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

pipe = pipeline("text-generation", model="PIPer-iclr/PIPer-8B-RL-only")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("PIPer-iclr/PIPer-8B-RL-only")
model = AutoModelForCausalLM.from_pretrained("PIPer-iclr/PIPer-8B-RL-only", 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 PIPer-iclr/PIPer-8B-RL-only with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PIPer-iclr/PIPer-8B-RL-only"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PIPer-iclr/PIPer-8B-RL-only",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/PIPer-iclr/PIPer-8B-RL-only

SGLang

How to use PIPer-iclr/PIPer-8B-RL-only 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 "PIPer-iclr/PIPer-8B-RL-only" \
    --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": "PIPer-iclr/PIPer-8B-RL-only",
		"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 "PIPer-iclr/PIPer-8B-RL-only" \
        --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": "PIPer-iclr/PIPer-8B-RL-only",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use PIPer-iclr/PIPer-8B-RL-only with Docker Model Runner:
```
docker model run hf.co/PIPer-iclr/PIPer-8B-RL-only
```

PIPer-8B-RL-only / README.md

PIPer-iclr

Update README.md

61d0b97 verified 11 months ago

preview code

Raw

History Blame Contribute Delete

3.32 kB

	---
	library_name: transformers
	datasets:
	- PIPer-iclr/envbench-zeroshot-rl
	base_model:
	- PIPer-iclr/Qwen3-8B-am
	pipeline_tag: text-generation
	license: mit
	---
	# 🚀 PIPer: On-Device Environment Setup via Online Reinforcement Learning

	<div align="center">

	[![Models](https://img.shields.io/badge/🤗%20Hugging%20Face-Models-orange.svg)](https://huggingface.co/PIPer-iclr)
	[![Dataset](https://img.shields.io/badge/🤗%20Hugging%20Face-Dataset-green.svg)](https://huggingface.co/datasets/PIPer-iclr/envbench-zeroshot-rl)
	[![License](https://img.shields.io/badge/License-MIT-red.svg)](LICENSE)

	Democratizing environment setup with on-device sized models that match the performance of much larger proprietary systems

	</div>

	## 🎯 Overview

	Environment setup—the process of configuring systems to work with specific software projects—remains a persistent challenge in software engineering. PIPer addresses this by training specialized on-device models that can automatically generate correct Bash scripts for environment configuration.

	Our approach combines:
	- 📚 Supervised Fine-Tuning (SFT) with executable scripts from larger models
	- 🎯 Reinforcement Learning with Verifiable Rewards (RLVR) using lightweight proxy LLM-reward

	## 🏆 Key Results

	\| Model \| Size \| EnvBench avg@5 \| Cost per 1M tokens \|
	\|-------\|------\|----------------\|-------------------\|
	\| PIPer \| 8B \| 19.4 \| $0.60 \|
	\| GPT-4o \| - \| 19.4 \| $15.00 \|
	\| Qwen3-32B \| 32B \| 16.2 \| $2.00 \|
	\| Qwen3-8B \| 8B \| 2.6 \| $0.60 \|

	> 🎉 PIPer achieves 9× improvement over its base model while matching GPT-4o performance at 25x lower cost

	![Performance vs Cost Analysis](https://github.com/PIPer-iclr/PIPer/raw/main/misc/combined_pass_n_and_cost.png)

	## 📦 Available Artifacts

	### 🤖 Model Checkpoints

	\| Model \| Description \| HuggingFace Link \|
	\|-------\|-------------\|------------------\|
	\| 🏅 PIPer (Full) \| Complete SFT+RL trained model \| [PIPer-iclr/PIPer-8B](https://huggingface.co/PIPer-iclr/PIPer-8B) \|
	\| 🎯 PIPer (RL-only) \| RLVR checkpoint only \| [PIPer-iclr/PIPer-8B-RL-only](https://huggingface.co/PIPer-iclr/PIPer-8B-RL-only) \|
	\| 📚 PIPer (SFT-only) \| Supervised fine-tuning only \| [PIPer-iclr/PIPer-8B-SFT-only](https://huggingface.co/PIPer-iclr/PIPer-8B-SFT-only) \|

	### 📊 Datasets

	\| Dataset \| Description \| HuggingFace Link \|
	\|---------\|-------------\|------------------\|
	\| EnvBench Zero-shot RL \| Training prompts and evaluation data \| [PIPer-iclr/envbench-zeroshot-rl](https://huggingface.co/datasets/PIPer-iclr/envbench-zeroshot-rl) \|

	## 📊 Evaluation Benchmarks

	\| Benchmark \| Description \| Metric \| Our Result \|
	\|-----------\|-------------\|---------\|------------\|
	\| EnvBench-Python \| 329 Python repositories \| pass@5 \| 🏆 27/329 \|
	\| Repo2Run \| 420 Python repositories \| pass@5 \| 🏆 103/420 \|
	\| Terminal-Bench \| 80 terminal tasks \| pass@10 \| 4/80 \|



	## 🙏 Acknowledgments

	- Built on top of [EnvBench](https://github.com/princeton-nlp/EnvBench) evaluation framework
	- Uses [VeRL](https://github.com/volcengine/verl) for efficient RL training
	- Leverages [Qwen3](https://huggingface.co/Qwen) model family as base architecture

	## 📄 License

	This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.