Instructions to use AINovice2005/quantized-SmolLM3-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AINovice2005/quantized-SmolLM3-3B with Transformers:

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

pipe = pipeline("text-generation", model="AINovice2005/quantized-SmolLM3-3B")
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("AINovice2005/quantized-SmolLM3-3B")
model = AutoModelForCausalLM.from_pretrained("AINovice2005/quantized-SmolLM3-3B", 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 AINovice2005/quantized-SmolLM3-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AINovice2005/quantized-SmolLM3-3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AINovice2005/quantized-SmolLM3-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AINovice2005/quantized-SmolLM3-3B

SGLang

How to use AINovice2005/quantized-SmolLM3-3B 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 "AINovice2005/quantized-SmolLM3-3B" \
    --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": "AINovice2005/quantized-SmolLM3-3B",
		"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 "AINovice2005/quantized-SmolLM3-3B" \
        --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": "AINovice2005/quantized-SmolLM3-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AINovice2005/quantized-SmolLM3-3B with Docker Model Runner:
```
docker model run hf.co/AINovice2005/quantized-SmolLM3-3B
```

Configuration Parsing Warning:In config.json: "quantization_config.modules_to_not_convert" must be an array

SmolLM3‑3B • Quantized

🚀 Model Description

This is an int8 quantized version of SmolLM3–3B, a highly efficient, open-source 3 B parameter LLM. It delivers nearly state-of-the-art multilingual reasoning and long-context performance (up to 128k tokens) with drastically reduced memory usage and inference cost, enabling fast deployment on mid‑range GPUs and edge devices.

📏 Quantization Details

Library: torchao
Precision: int8 weights and activations
Benefits: ~50–75% reduction in VRAM usage, enabling 12–16 GB GPU usage, with minimal performance drop on reasoning, coding, and long-context abilities

🎯 Intended Use

Ideal for:

Scenarios requiring fast LLM inference under constrained VRAM (e.g. small servers or laptops)
Multilingual reasoning tasks, chain-of-thought logic, and long-context document understanding
Deployments of dual-mode (think/no_think) conversational agents
Research into efficient LLM deployment and quantization techniques

⚠️ Limitations

Slight performance loss compared to full-precision SmolLM3‑3B
Requires proper benchmarking in your actual environment
Continues to exhibit standard LLM risks: hallucination, bias, etc.
Quant performance may vary across languages or context lengths

Downloads last month: 7

Model tree for AINovice2005/quantized-SmolLM3-3B

Base model

HuggingFaceTB/SmolLM3-3B-Base

Finetuned

HuggingFaceTB/SmolLM3-3B

Quantized

(131)

this model

Collection including AINovice2005/quantized-SmolLM3-3B

🚀 Optimized Models: torchao & Pruna Quantization

Collection

Quantized Models using torchao & Pruna for efficient inference and deployment. • 8 items • Updated Mar 12 • 1