Intel/gpt-oss-20b-int4-rtn-AutoRound

Model Details

This model is a mixed int4 model with group_size 128 and symmetric quantization of openai/gpt-oss-20b generated by intel/auto-round via RTN (no algorithm tuning). Following the official model, Non expert layers fallback to 16bits.

How To Use

this model could not run on vLLM

INT4 Inference on CPU/Intel GPU/CUDA

from transformers import pipeline

model_id ="Intel/gpt-oss-20b-int4-rtn-AutoRound"

pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True
)

messages = [
    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
]

outputs = pipe(
    messages,
    max_new_tokens=512,
)
print(outputs[0]["generated_text"][-1])

"""
{'role': 'assistant', 'content': "analysisUser wants a clear and concise explanation of quantum mechanics. Likely want key concepts: wave-particle duality, superposition, uncertainty principle, quantum states, measurement, entanglement, etc. Keep it concise but thorough. Probably 3-4 paragraphs. Let's produce.assistantfinal**Quantum mechanics** is the physics of the very small—atoms, electrons, photons, and other elementary particles. It replaces the deterministic, continuous description of classical mechanics with a probabilistic, discrete one. The key ideas are:\n\n| Concept | What it means | Why it matters |\n|--------|----------------|----------------|\n| **Wave‑particle duality** | Particles such as electrons behave both like particles (having a definite position when measured) and like waves (exhibiting interference). | Explains phenomena like the double‑slit experiment, where particles create an interference pattern unless we look at them. |\n| **Quantum state & wavefunction** | The state of a system is described by a complex wavefunction ψ(x). Its square gives a probability density | The wavefunction contains all we can know about a system; it evolves deterministically via Schrödinger’s equation. |\n| **Superposition** | A system can be in a"}

""

Ethical Considerations and Limitations

The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs.

Therefore, before deploying any applications of the model, developers should perform safety testing.

Caveats and Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.

Here are a couple of useful links to learn more about Intel's AI software:

• Intel Neural Compressor link

Disclaimer

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes.

Cite

@article{cheng2023optimize, title={Optimize weight rounding via signed gradient descent for the quantization of llms}, author={Cheng, Wenhua and Zhang, Weiwei and Shen, Haihao and Cai, Yiyang and He, Xin and Lv, Kaokao and Liu, Yi}, journal={arXiv preprint arXiv:2309.05516}, year={2023} }

arxiv github