Instructions to use oracomputing/Qwen3-4B-ORA-W3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use oracomputing/Qwen3-4B-ORA-W3 with Transformers:

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

pipe = pipeline("text-generation", model="oracomputing/Qwen3-4B-ORA-W3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("oracomputing/Qwen3-4B-ORA-W3")
model = AutoModelForCausalLM.from_pretrained("oracomputing/Qwen3-4B-ORA-W3", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use oracomputing/Qwen3-4B-ORA-W3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "oracomputing/Qwen3-4B-ORA-W3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "oracomputing/Qwen3-4B-ORA-W3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/oracomputing/Qwen3-4B-ORA-W3

SGLang

How to use oracomputing/Qwen3-4B-ORA-W3 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 "oracomputing/Qwen3-4B-ORA-W3" \
    --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": "oracomputing/Qwen3-4B-ORA-W3",
		"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 "oracomputing/Qwen3-4B-ORA-W3" \
        --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": "oracomputing/Qwen3-4B-ORA-W3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use oracomputing/Qwen3-4B-ORA-W3 with Docker Model Runner:
```
docker model run hf.co/oracomputing/Qwen3-4B-ORA-W3
```

rmedinar commited on 4 days ago

Commit

acbe254

verified ·

1 Parent(s): a5276f6

Upload folder using huggingface_hub

Browse files

Files changed (1) hide show

README.md +2 -2

README.md CHANGED Viewed

@@ -18,7 +18,7 @@ tags:
   <br/>
   <a href="https://www.oracomputing.com/en"><b>Website</b></a>
   &nbsp;·&nbsp;
-  <a href="https://www.oracomputing.com/en/blog"><b>Blog</b></a>
   &nbsp;·&nbsp;
   <a href="https://www.oracomputing.com/en/contact"><b>Contact</b></a>
 </div>
@@ -28,7 +28,7 @@ tags:
 > [!IMPORTANT]
 > **~3.7× smaller** than the original 16-bit [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B), with **96.5% accuracy retention**.
-3-bit weight-only quantization for [Qwen/Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B) using our propietary Qauntization-Aware-Training pipeline, more information in the dedicated [post](https://www.oracomputing.com/en/blog).
 **Serve with vLLM ≥ 0.25.0** (Humming WNA16) to keep weights packed. Transformers also works if you pin `compressed-tensors>=0.18` but it decompresses the 3-bit weights to bf16 in memory.

   <br/>
   <a href="https://www.oracomputing.com/en"><b>Website</b></a>
   &nbsp;·&nbsp;
+  <a href="https://www.oracomputing.com/en/blog/ora-qat"><b>Blog</b></a>
   &nbsp;·&nbsp;
   <a href="https://www.oracomputing.com/en/contact"><b>Contact</b></a>
 </div>
 > [!IMPORTANT]
 > **~3.7× smaller** than the original 16-bit [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B), with **96.5% accuracy retention**.
+3-bit weight-only quantization for [Qwen/Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B) using our propietary Qauntization-Aware-Training pipeline, more information in the dedicated [post](https://www.oracomputing.com/en/blog/ora-qat).
 **Serve with vLLM ≥ 0.25.0** (Humming WNA16) to keep weights packed. Transformers also works if you pin `compressed-tensors>=0.18` but it decompresses the 3-bit weights to bf16 in memory.