Instructions to use s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6 with Transformers:

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

pipe = pipeline("text-generation", model="s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6")
model = AutoModelForCausalLM.from_pretrained("s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6", 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 s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6

SGLang

How to use s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6 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 "s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6" \
    --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": "s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6",
		"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 "s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6" \
        --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": "s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6 with Docker Model Runner:
```
docker model run hf.co/s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6
```

Countdown-Qwen2.5-0.5B-GRPO (lr 3e-6 ablation)

Qwen2.5-0.5B-Instruct fine-tuned with from-scratch GRPO (Group-Relative Policy Optimization) on the Countdown number-puzzle task — combine the given numbers exactly once with + - * / to hit a target. Rewards are verifiable (exact rational arithmetic via a whitelisted-AST evaluator), so this is RLVR: no reward model, no critic.

This checkpoint is the best ablation (learning rate 3e-6, group 8, 1500 steps).

Results (dev_public, 300 puzzles, greedy, exact verifier)

model	accuracy	hard (5-num)	format_rate	avg_tokens (correct)
base Qwen2.5-0.5B-Instruct (floor)	0.33%	0.00%	0.00%	20.0
this model (GRPO, lr 3e-6)	12.00%	1.67%	0.00%	16.9

36x over the base floor (1/300 -> 36/300). Ablations: learning rate was the only lever that moved accuracy (1e-6 -> 3e-6 gave 1% -> 12%); more steps (3000) and larger group (16) both plateaued. Known failure mode: reasoning collapse — the shaped reward scores only the <answer>, so the model emits bare answers with no <think> (format_rate = 0).

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("<your-username>/countdown-qwen2.5-0.5b-grpo-lr3e6")
tok = AutoTokenizer.from_pretrained("<your-username>/countdown-qwen2.5-0.5b-grpo-lr3e6")

Trained for the RLVR Arena capstone (RL in Production Bootcamp).

Downloads last month: 23

Safetensors

Model size

0.5B params

Tensor type

BF16

Model tree for s1lv3rj1nx/countdown-qwen2.5-0.5b-grpo-lr3e6

Base model

Qwen/Qwen2.5-0.5B

Finetuned

Qwen/Qwen2.5-0.5B-Instruct

Finetuned

(1013)

this model