Instructions to use siliconcorerina/rina-coder-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use siliconcorerina/rina-coder-base with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="siliconcorerina/rina-coder-base")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("siliconcorerina/rina-coder-base", dtype="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use siliconcorerina/rina-coder-base with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "siliconcorerina/rina-coder-base" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "siliconcorerina/rina-coder-base", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/siliconcorerina/rina-coder-base
- SGLang
How to use siliconcorerina/rina-coder-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 "siliconcorerina/rina-coder-base" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "siliconcorerina/rina-coder-base", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'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 "siliconcorerina/rina-coder-base" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "siliconcorerina/rina-coder-base", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use siliconcorerina/rina-coder-base with Docker Model Runner:
docker model run hf.co/siliconcorerina/rina-coder-base
RINA Coder โ Base
Modele de langage RINA AI dedie a la generation, la completion et l explication de code. Site : plateforme-rina.com ยท Code : github.com/siliconcorerina/RINA-AI
Statut : placeholder. Les poids ne sont pas encore publies. Ce depot reserve l identifiant siliconcorerina/rina-coder-base et decrit le modele cible. La premiere version sera annoncee via les issues du depot GitHub.
Description
RINA Coder est la famille de modeles de generation de code maintenue par l equipe RINA AI. Cette variante base est destinee a la completion et a la generation libre. Une variante instruct suivra pour les usages conversationnels.
Usage prevu
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "siliconcorerina/rina-coder-base"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
prompt = "def fibonacci(n):"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=128, temperature=0.2)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Voir aussi le script de demo : demo/inference_example.py.
Cas d usage
- Completion de code dans des editeurs et IDE
- Generation de fonctions a partir de docstrings
- Explication de snippets de code
- Refactoring assiste
- Tests unitaires generes a partir du code source
Hors perimetre
- Conseil juridique, medical ou financier
- Decisions impactant des personnes (recrutement, credit, etc.)
- Usage en production sans verification humaine du code genere
Donnees d entrainement
A documenter lors de la publication du premier checkpoint. Les sources prevues incluent :
- Code open source sous licences permissives
- Documentation technique publique
- Corpus de problemes de programmation (HumanEval-like)
Evaluation
Les benchmarks cibles sont :
| Benchmark | Statut |
|---|---|
| HumanEval (pass@1) | a venir |
| MBPP (pass@1) | a venir |
| MultiPL-E (Rust, Go, Kotlin) | a venir |
| RINA-Bench (interne) | a venir |
Suivi : issues evaluation.
Limitations
- Le code genere peut contenir des bugs, des failles de securite, ou ne pas compiler. Toujours relire et tester.
- Le modele peut halluciner des API ou des bibliotheques inexistantes.
- Les performances varient fortement selon le langage et le domaine.
- Le contexte est limite ; les fichiers tres longs ne sont pas couverts dans une seule passe.
Licence
MIT. Voir LICENSE.
Contact
- Site : plateforme-rina.com
- Email : hello@plateforme-rina.com
- GitHub : github.com/siliconcorerina
Citation
@misc{rinacoder2026,
title = {RINA Coder: a code language model by RINA AI},
author = {RINA AI Team},
year = {2026},
url = {https://huggingface.co/siliconcorerina/rina-coder-base}
}