Instructions to use jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning with Transformers:

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

pipe = pipeline("text-generation", model="jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning")
model = AutoModelForCausalLM.from_pretrained("jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning")
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

vLLM

How to use jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning

SGLang

How to use jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning 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 "jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning" \
    --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": "jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning",
		"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 "jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning" \
        --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": "jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning with Docker Model Runner:
```
docker model run hf.co/jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning
```

Qwen2.5-0.5B-GRPO-KL-math-reasoning

This model is a fine-tuned version of Qwen2.5-0.5B using GRPO (Group Relative Policy Optimization) with KL penalty for mathematical reasoning.

Trained with PipelineRL.

Training Details

Datasets

Split	Datasets
Train	`gsm8k_train`, `math_train`
Test	`gsm8k_test`, `math_500`

RL Algorithm

Parameter	Value
Algorithm	GRPO (Group Relative Policy Optimization)
Advantage Baseline	Group mean reward
Extra Inference	None
Group Structure	Required
Policy Loss	`ppo`
KL Coefficient	`0.001`
Epsilon (clip)	`0.02`
Discount Factor (`gamma`)	`1.0`
Divide Advantage by Std	`False`
Filter Zero Advantage Groups	`False`
Rollouts per Problem	`16`

GRPO uses the group mean reward as the baseline for relative advantages.

Training Hyperparameters

Parameter	Value
Base Model	`Qwen/Qwen2.5-0.5B`
Learning Rate	`1e-06`
LR Scheduler	`cosine`
Warmup Steps	`25`
Max Training Steps	`1500`
Micro Batch Size	`8`
Gradient Accumulation	`32`
Effective Batch Size	`256`
Sequence Length	`8192`
Gradient Clipping	`0.3`
Weight Decay	`0.01`
Optimizer	`adamw_torch`
Precision	`bf16`
DeepSpeed	ZeRO Stage 3

Evaluation Results

Pass@k on math reasoning benchmarks (N=32 samples per problem, temperature=1.0):

Dataset	pass@1	pass@2	pass@4	pass@8	pass@16	pass@32
GSM8K (test)	48.67	59.36	68.41	76.01	82.52	87.79
MATH-500	30.31	39.38	48.14	56.47	64.04	70.40
Overall	43.62	53.87	62.84	70.64	77.44	83.01

GSM8K test: 1319 problems · MATH-500: 500 problems · Overall: 1819 problems (overall weighted by problem count).

Training Curves

$Training Metrics$

W&B Run

Full training logs: https://wandb.ai/jaygala24-team/rl-post-training/runs/qwen2.5_0.5b_grpo_with_kl_2a1p1f_4xh100_202889_finetune_e04844e9

Usage

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning", revision="step-0200")  # optional branch, e.g. "step-0400"
tokenizer = AutoTokenizer.from_pretrained("jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning", revision="step-0200")

prompt = "Please reason step by step, and put your final answer within \\boxed{}.\n\nWhat is the sum of 123 and 456?"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=4096, temperature=0.7)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

vLLM

from vllm import LLM, SamplingParams

llm = LLM(model="jaygala24/Qwen2.5-0.5B-GRPO-KL-math-reasoning", revision="step-0200")  # optional branch, e.g. "step-0400"
sampling_params = SamplingParams(temperature=0.7, max_tokens=4096)

prompt = "Please reason step by step, and put your final answer within \boxed{}.

What is the sum of 123 and 456?"
outputs = llm.generate([prompt], sampling_params)
print(outputs[0].outputs[0].text)