Instructions to use krogoldAI/QueryRefiner-0.5B-v0.1-GRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use krogoldAI/QueryRefiner-0.5B-v0.1-GRPO with Transformers:

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

pipe = pipeline("text-generation", model="krogoldAI/QueryRefiner-0.5B-v0.1-GRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("krogoldAI/QueryRefiner-0.5B-v0.1-GRPO")
model = AutoModelForCausalLM.from_pretrained("krogoldAI/QueryRefiner-0.5B-v0.1-GRPO", 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 krogoldAI/QueryRefiner-0.5B-v0.1-GRPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "krogoldAI/QueryRefiner-0.5B-v0.1-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": "krogoldAI/QueryRefiner-0.5B-v0.1-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/krogoldAI/QueryRefiner-0.5B-v0.1-GRPO

SGLang

How to use krogoldAI/QueryRefiner-0.5B-v0.1-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 "krogoldAI/QueryRefiner-0.5B-v0.1-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": "krogoldAI/QueryRefiner-0.5B-v0.1-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 "krogoldAI/QueryRefiner-0.5B-v0.1-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": "krogoldAI/QueryRefiner-0.5B-v0.1-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use krogoldAI/QueryRefiner-0.5B-v0.1-GRPO with Docker Model Runner:
```
docker model run hf.co/krogoldAI/QueryRefiner-0.5B-v0.1-GRPO
```

krogoldAI commited on Oct 12, 2025

Commit

b61271b

verified ·

1 Parent(s): 3c35203

Update README.md

Browse files

Files changed (1) hide show

README.md +3 -3

README.md CHANGED Viewed

@@ -150,15 +150,15 @@ This stratified distribution ensures the model learns to handle the full spectru
 QueryRefiner-0.5B-v0.1-GRPO underwent a sophisticated three-phase training procedure combining reinforcement learning with two-stage supervised fine-tuning, all using full parameter updates (not parameter-efficient methods like LoRA) on [Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct).
 ### Phase 1: Reinforcement Learning with GRPO
-The model was first trained using Group Relative Policy Optimization (GRPO) on 3,000 examples from [krogoldAI/rag-ambiguous-queries](https://huggingface.co/datasets/krogoldAI/rag-ambiguous-queries) for 1 epoch with a learning rate of 5e-6 (intentionally low for GRPO stability). This phase focused on learning correct XML structure and formatting before semantic refinement.
-The GRPO reward function evaluated outputs through a weighted combination of five components: tag structure (30%), XML validity (25%), element ordering (25%), confidence formatting (18%), and confidence distribution (2%). The structure component verified the presence of all required XML elements, while validity ensured parseability. The ordering component checked that tags appeared in the correct sequence, and the confidence component validated that confidence values were properly formatted and summed to 1.0 for ambiguous cases.
 The confidence distribution component used normalized entropy to encourage balanced probability distributions in ambiguous classifications, computed as:
 $$ -\frac{1}{\log n}\sum_{j=1}^n p_j \log(p_j+\delta) $$
-where \\( (p_j)_{1\leq j\leq n} \\) are the confidence scores, \\( n \\) is the number of confidence scores, \\( \log n \\) represents the maximum possible entropy and \\( \delta = 10^{-8} \\) is a small regularization constant. This penalizes overly skewed distributions (e.g., 0.9/0.1 splits) in favor of more balanced confidence allocations when multiple interpretations are plausible. Additional penalties were applied for structural issues (missing attributes) and for outputs that attempted to answer queries rather than analyze them.
 ### Phase 2: Breadth-Focused Supervised Fine-Tuning
 Following structural learning through GRPO, the model underwent supervised fine-tuning on the complete [krogoldAI/rag-query-analysis](https://huggingface.co/datasets/krogoldAI/rag-query-analysis) dataset (7,305 examples) for 2 epochs with a learning rate of 2e-5. This phase captured broad semantic patterns across diverse query types and domains.

 QueryRefiner-0.5B-v0.1-GRPO underwent a sophisticated three-phase training procedure combining reinforcement learning with two-stage supervised fine-tuning, all using full parameter updates (not parameter-efficient methods like LoRA) on [Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct).
 ### Phase 1: Reinforcement Learning with GRPO
+The model was first trained using Group Relative Policy Optimization (GRPO) on 3,000 examples from [krogoldAI/rag-ambiguous-queries](https://huggingface.co/datasets/krogoldAI/rag-ambiguous-queries) for 1 epoch with a learning rate of 5e-6 (low for GRPO stability). This phase focused on learning correct XML structure and formatting before semantic refinement.
+The GRPO reward function evaluated outputs through a weighted combination of five components: tag structure (30%), XML validity (25%), element ordering (25%), confidence formatting (18%), and confidence distribution (2%). The structure component verified the presence of all required XML elements, while validity ensured parseability. The ordering component checked that tags appeared in the correct sequence, and the confidence component validated that confidence values were properly formatted and summed to `1.0` for ambiguous cases.
 The confidence distribution component used normalized entropy to encourage balanced probability distributions in ambiguous classifications, computed as:
 $$ -\frac{1}{\log n}\sum_{j=1}^n p_j \log(p_j+\delta) $$
+where \\( (p_j)_{1\leq j\leq n} \\) are the confidence scores, \\( n \\) is the number of confidence scores, \\( \log n \\) is the maximum possible entropy and \\( \delta = 10^{-8} \\) is a small regularization constant. This penalizes overly skewed distributions (e.g., `0.9`/`0.1` splits) in favor of more balanced confidence allocations when multiple interpretations are plausible. Additional penalties were applied for structural issues (missing attributes) and for outputs that attempted to answer queries rather than analyze them.
 ### Phase 2: Breadth-Focused Supervised Fine-Tuning
 Following structural learning through GRPO, the model underwent supervised fine-tuning on the complete [krogoldAI/rag-query-analysis](https://huggingface.co/datasets/krogoldAI/rag-query-analysis) dataset (7,305 examples) for 2 epochs with a learning rate of 2e-5. This phase captured broad semantic patterns across diverse query types and domains.