Instructions to use dev-bjoern/smollm3-int4-ov with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dev-bjoern/smollm3-int4-ov with Transformers:

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

pipe = pipeline("text-generation", model="dev-bjoern/smollm3-int4-ov")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dev-bjoern/smollm3-int4-ov")
model = AutoModelForCausalLM.from_pretrained("dev-bjoern/smollm3-int4-ov")
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 dev-bjoern/smollm3-int4-ov with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dev-bjoern/smollm3-int4-ov"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dev-bjoern/smollm3-int4-ov",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dev-bjoern/smollm3-int4-ov

SGLang

How to use dev-bjoern/smollm3-int4-ov 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 "dev-bjoern/smollm3-int4-ov" \
    --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": "dev-bjoern/smollm3-int4-ov",
		"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 "dev-bjoern/smollm3-int4-ov" \
        --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": "dev-bjoern/smollm3-int4-ov",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use dev-bjoern/smollm3-int4-ov with Docker Model Runner:
```
docker model run hf.co/dev-bjoern/smollm3-int4-ov
```

dev-bjoern commited on Aug 6, 2025

Commit

29f8c2d

verified ·

1 Parent(s): 00bdc52

Update README.md

Browse files

Files changed (1) hide show

README.md +11 -8

README.md CHANGED Viewed

@@ -30,9 +30,9 @@ This is an INT4 quantized version of [SmolLM3-3B](https://huggingface.co/Hugging
 ## Model Overview
-- **Base Model:** SmolLM3-3B (3B parameters)
 - **Quantization:** INT4 via OpenVINO
-- **Size Reduction:** ~75% smaller than original
 - **Target Hardware:** CPUs, Intel GPUs, NPUs
 - **Use Cases:** Local inference, edge deployment, resource-constrained environments
@@ -55,10 +55,13 @@ This is an INT4 quantized version of [SmolLM3-3B](https://huggingface.co/Hugging
 > ⚠️ **Note:** This is an experimental quantization. Formal benchmarks pending.
-Expected characteristics:
-- **Model Size:** ~1GB (vs ~6GB fp16)
-- **Inference Speed:** 2-4x faster on CPU
-- **Quality Trade-off:** Minor degradation expected
 ## 🛠️ How to Use
@@ -108,8 +111,8 @@ text = tokenizer.apply_chat_template(
 ## ⚡ Optimization Tips
 1. **CPU Inference:** Use OpenVINO runtime for best performance
-2. **Batch Processing:** Leverage dynamic batching when possible
-3. **Memory:** Requires ~2GB RAM for comfortable operation
 ## 🧪 Experimental Status

 ## Model Overview
+- **Base Model:** SmolLM3-3B
 - **Quantization:** INT4 via OpenVINO
+- **Size Reduction:** Significant compression achieved
 - **Target Hardware:** CPUs, Intel GPUs, NPUs
 - **Use Cases:** Local inference, edge deployment, resource-constrained environments
 > ⚠️ **Note:** This is an experimental quantization. Formal benchmarks pending.
+Expected benefits of INT4 quantization:
+- Reduced model size
+- Faster CPU inference
+- Lower memory requirements
+- Some quality trade-off
+Actual metrics will be added after proper benchmarking.
 ## 🛠️ How to Use
 ## ⚡ Optimization Tips
 1. **CPU Inference:** Use OpenVINO runtime for best performance
+2. **Batch Processing:** Consider batching requests when possible
+3. **Memory:** INT4 significantly reduces memory requirements
 ## 🧪 Experimental Status