Instructions to use Praneshrajan15/DataForge-0.5B-GRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Praneshrajan15/DataForge-0.5B-GRPO with Transformers:

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

pipe = pipeline("text-generation", model="Praneshrajan15/DataForge-0.5B-GRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("Praneshrajan15/DataForge-0.5B-GRPO")
model = AutoModelForMultimodalLM.from_pretrained("Praneshrajan15/DataForge-0.5B-GRPO")
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 Praneshrajan15/DataForge-0.5B-GRPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Praneshrajan15/DataForge-0.5B-GRPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Praneshrajan15/DataForge-0.5B-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Praneshrajan15/DataForge-0.5B-GRPO

SGLang

How to use Praneshrajan15/DataForge-0.5B-GRPO 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 "Praneshrajan15/DataForge-0.5B-GRPO" \
    --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": "Praneshrajan15/DataForge-0.5B-GRPO",
		"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 "Praneshrajan15/DataForge-0.5B-GRPO" \
        --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": "Praneshrajan15/DataForge-0.5B-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Praneshrajan15/DataForge-0.5B-GRPO with Docker Model Runner:
```
docker model run hf.co/Praneshrajan15/DataForge-0.5B-GRPO
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

DataForge-0.5B-GRPO

GRPO checkpoint trained from predecessor Praneshrajan15/DataForge-0.5B-SFT and uploaded only after the strict held-out gate passed with F1 delta 0.134.

Evidence

Benchmark: DataForge-Bench-light-verified over seeds [0, 1, 2]
SFT strict macro F1: 0.0053
GRPO strict macro F1: 0.1393
Parse success: 1.0
Schema-case errors: 0
Training stage: candidate with 500 steps

Limitations

This is a research artifact for DataForge repair evaluation. It is not production autonomous repair software and must not mutate data without DataForge verification, receipts, and human approval.

Downloads last month: 17

Safetensors

Model size

0.5B params

Tensor type

F16

Model tree for Praneshrajan15/DataForge-0.5B-GRPO

Base model

Qwen/Qwen2.5-0.5B

Finetuned

Qwen/Qwen2.5-0.5B-Instruct

Finetuned

(822)

this model

Quantizations

1 model

Dataset used to train Praneshrajan15/DataForge-0.5B-GRPO

Evaluation results

Strict macro F1 on DataForge-Bench-light-verified
self-reported

0.139
Delta over predecessor on DataForge-Bench-light-verified
self-reported

0.134