Instructions to use vedssharma/Qwen3-1.7B-Base-MBPP-GRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vedssharma/Qwen3-1.7B-Base-MBPP-GRPO with Transformers:

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

pipe = pipeline("text-generation", model="vedssharma/Qwen3-1.7B-Base-MBPP-GRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("vedssharma/Qwen3-1.7B-Base-MBPP-GRPO")
model = AutoModelForCausalLM.from_pretrained("vedssharma/Qwen3-1.7B-Base-MBPP-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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use vedssharma/Qwen3-1.7B-Base-MBPP-GRPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vedssharma/Qwen3-1.7B-Base-MBPP-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": "vedssharma/Qwen3-1.7B-Base-MBPP-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/vedssharma/Qwen3-1.7B-Base-MBPP-GRPO

SGLang

How to use vedssharma/Qwen3-1.7B-Base-MBPP-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 "vedssharma/Qwen3-1.7B-Base-MBPP-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": "vedssharma/Qwen3-1.7B-Base-MBPP-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 "vedssharma/Qwen3-1.7B-Base-MBPP-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": "vedssharma/Qwen3-1.7B-Base-MBPP-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use vedssharma/Qwen3-1.7B-Base-MBPP-GRPO with Docker Model Runner:
```
docker model run hf.co/vedssharma/Qwen3-1.7B-Base-MBPP-GRPO
```

Qwen3-1.7B-Base-MBPP-GRPO

Qwen3-1.7B-Base finetuned with GRPO (TRL + vLLM, single H100 in a Daytona sandbox) on ~460 MBPP problems (train+validation+prompt splits). Reward: 1 if the generated code passes all unit tests, else 0.

200 steps, lr 1e-6, 8 prompts x 8 samples per step, 512 max completion tokens, no KL term.

Model	MBPP test pass@1 (greedy, n=500)
Qwen3-1.7B-Base	52.0%
This model	57.0%

Prompt format: a plain-text problem plus first test, ending with an opened "```python" fence; generation stops at the closing fence. Small-scale experiment; the training set is small and gains may not generalize beyond MBPP-style tasks. Weights are fp32 as saved by the trainer.

Downloads last month: 133

Safetensors

Model size

2B params

Tensor type

F32

Model tree for vedssharma/Qwen3-1.7B-Base-MBPP-GRPO

Base model

Qwen/Qwen3-1.7B-Base

Finetuned

(503)

this model