Instructions to use cazonai/elyra-0.1-gemma3-4b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use cazonai/elyra-0.1-gemma3-4b with Transformers:

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

pipe = pipeline("text-generation", model="cazonai/elyra-0.1-gemma3-4b")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("cazonai/elyra-0.1-gemma3-4b")
model = AutoModelForImageTextToText.from_pretrained("cazonai/elyra-0.1-gemma3-4b")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use cazonai/elyra-0.1-gemma3-4b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "cazonai/elyra-0.1-gemma3-4b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cazonai/elyra-0.1-gemma3-4b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/cazonai/elyra-0.1-gemma3-4b

SGLang

How to use cazonai/elyra-0.1-gemma3-4b 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 "cazonai/elyra-0.1-gemma3-4b" \
    --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": "cazonai/elyra-0.1-gemma3-4b",
		"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 "cazonai/elyra-0.1-gemma3-4b" \
        --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": "cazonai/elyra-0.1-gemma3-4b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use cazonai/elyra-0.1-gemma3-4b with Docker Model Runner:
```
docker model run hf.co/cazonai/elyra-0.1-gemma3-4b
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Elyra 0.1 — A IA da Cazon AI

Elyra é a LLM oficial da Cazon AI (Renan Cazon, Brodowski/SP, Brasil), v0.1.

Fine-tuned do Gemma 3 4B via Unsloth QLoRA em 9.202 amostras destiladas focadas em:

Identidade Cazon AI
Produto AutoRadio (dongle ESP32+GPS+4G de troca automática de rádio FM)
Rádios brasileiras (~3000 emissoras, foco interior SP)
Cultura do motorista brasileiro / caminhoneiro
Português brasileiro informal e direto

Treino

Base: unsloth/gemma-3-4b-it
Método: LoRA (r=16, alpha=32) em 7 módulos
Dataset: cazonai/autoradio-destilado-v1 (9.202 samples)
Epochs: 3
Train loss final: 0.1949
GPU: NVIDIA A100 80GB PCIe
Tempo de treino: 2h 36min
Trainable params: 32.8M / 4.3B (0.76%)

System prompt padrão

Você é a ELYRA — a IA oficial da Cazon AI, criada por Renan Cazon (Brodowski, SP).
Responda em PT-BR informal e direto. Ao Renan, use "Maestro".
Nunca invente features do AutoRadio ou preços.

Limitações conhecidas (v0.1)

Conhece bem: AutoRadio, rádios BR, Brodowski, identidade Cazon AI
NÃO conhece (virá nas próximas versões): Projeto Resonance, ElyraOS Desktop, Cazon Bank