Instructions to use ssurface/qwen3-4b-gdpo-length-sft-l4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ssurface/qwen3-4b-gdpo-length-sft-l4 with Transformers:

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

pipe = pipeline("text-generation", model="ssurface/qwen3-4b-gdpo-length-sft-l4")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ssurface/qwen3-4b-gdpo-length-sft-l4")
model = AutoModelForCausalLM.from_pretrained("ssurface/qwen3-4b-gdpo-length-sft-l4", 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 ssurface/qwen3-4b-gdpo-length-sft-l4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ssurface/qwen3-4b-gdpo-length-sft-l4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ssurface/qwen3-4b-gdpo-length-sft-l4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ssurface/qwen3-4b-gdpo-length-sft-l4

SGLang

How to use ssurface/qwen3-4b-gdpo-length-sft-l4 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 "ssurface/qwen3-4b-gdpo-length-sft-l4" \
    --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": "ssurface/qwen3-4b-gdpo-length-sft-l4",
		"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 "ssurface/qwen3-4b-gdpo-length-sft-l4" \
        --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": "ssurface/qwen3-4b-gdpo-length-sft-l4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ssurface/qwen3-4b-gdpo-length-sft-l4 with Docker Model Runner:
```
docker model run hf.co/ssurface/qwen3-4b-gdpo-length-sft-l4
```

Qwen3-4B GRPO — Level 4 (Shorthand)

Qwen3-4B-Instruct fine-tuned with SFT → GRPO + new reward for compressed chain-of-thought reasoning at Level 4 (Shorthand).

Training Pipeline

Qwen/Qwen3-4B-Instruct-2507
  → SFT LoRA (ssurface/qwen3-4b-cot-compress-l4)
  → Merged
  → GRPO with new reward
  → Merged (this model)

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("ssurface/qwen3-4b-gdpo-length-sft-l4")
tokenizer = AutoTokenizer.from_pretrained("ssurface/qwen3-4b-gdpo-length-sft-l4")

messages = [{"role": "user", "content": "Solve this using Level 4 (Shorthand).\nProblem: ..."}]
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
outputs = model.generate(inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=False))

Downloads last month: 55

Safetensors

Model size

4B params

Tensor type

F16

Model tree for ssurface/qwen3-4b-gdpo-length-sft-l4

Base model

Qwen/Qwen3-4B-Instruct-2507

Finetuned

(1878)

this model

Collection including ssurface/qwen3-4b-gdpo-length-sft-l4

GRPO SFT-Length-Punishment-GDPO SFT GSM8K

Collection

5 items • Updated Jul 1