Instructions to use luminousresearch/L0-PolyCore-4B-Base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use luminousresearch/L0-PolyCore-4B-Base with Transformers:

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

pipe = pipeline("text-generation", model="luminousresearch/L0-PolyCore-4B-Base")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("luminousresearch/L0-PolyCore-4B-Base")
model = AutoModelForCausalLM.from_pretrained("luminousresearch/L0-PolyCore-4B-Base")
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 luminousresearch/L0-PolyCore-4B-Base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "luminousresearch/L0-PolyCore-4B-Base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "luminousresearch/L0-PolyCore-4B-Base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/luminousresearch/L0-PolyCore-4B-Base

SGLang

How to use luminousresearch/L0-PolyCore-4B-Base 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 "luminousresearch/L0-PolyCore-4B-Base" \
    --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": "luminousresearch/L0-PolyCore-4B-Base",
		"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 "luminousresearch/L0-PolyCore-4B-Base" \
        --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": "luminousresearch/L0-PolyCore-4B-Base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use luminousresearch/L0-PolyCore-4B-Base with Docker Model Runner:
```
docker model run hf.co/luminousresearch/L0-PolyCore-4B-Base
```

L0-PolyCore-4B-Base / README.md

r-e1

Update README.md

de560ca verified 3 months ago

preview code

raw

history blame contribute delete

1.35 kB

	---
	license: other
	language:
	- en
	base_model:
	- Qwen/Qwen3-4B-Base
	base_model_relation: finetune
	tags:
	- code
	- c
	- clang
	- cpp
	- c++
	- qlora
	- cpt
	library_name: transformers
	pipeline_tag: text-generation
	---
	## Training Data

	This model was trained on a dataset of curated C/C++ code from multiple licenses (GPL-2.0, Apache-2.0, MIT, public domain, and some source-available licenses, etc.).
	The original authors are not affiliated with or responsible for this model.

	## Base Model

	Base model: [Qwen/Qwen3-4B-Base](https://huggingface.co/Qwen/Qwen3-4B-Base)

	## Fine-tuning Method

	- Adapter: QLoRA
	- Method: CPT
	- Precision: trained with 4-bit base weights + BF16 compute, then merged to safetensors

	## Training Details

	- Training time: ~74 hours
	- Hardware: 1x NVIDIA RTX 5060 Ti

	## Notes

	- This is an L0 base model, it is not instruction-tuned and may be more verbose with strict formatting request compared to an instruct model.
	- Recommended usage is raw code continuation, or pairing with an external template strategy.

	## Intended use

	- Code generation for C/C++
	- Fast code completion
	- Examples and prototyping

	## Constraints

	- May produce incorrect code
	- May reproduce identifiable upstream code fragments (including license headers) when prompted.
	- Verify outputs, especially for memory safety and security-sensitive code.