namgyu-youn/Qwen3-8B-INT4

Note: This configs are specialized to low M (computation intensity; batch x SeqLen). It shows poor throughput for large M scenario. See https://github.com/pytorch/ao/issues/3496 for more info

INT4 Qwen/Qwen3-8B model

• Developed by: namgyu-youn
• License: apache-2.0
• Quantized from Model : Qwen/Qwen3-8B
• Quantization Method : INT4

Model Quality

We rely on lm-evaluation-harness to evaluate the quality of the quantized model. Here we only run on mmlu for sanity check.

Original

hf (pretrained=Qwen/Qwen3-8B), gen_kwargs: (None), limit: 200.0, num_fewshot: None, batch_size: 1

lm_eval --model hf     --model_args pretrained=Qwen/Qwen3-8B     --tasks hellaswag     --device cuda:0     --batch_size 1 --limit 200

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
hellaswag	1	none	0	acc	↑	0.490	±	0.0354
		none	0	acc_norm	↑	0.625	±	0.0343

Quantized (HQQ)

hf (pretrained=namgyu-youn/Qwen3-8B-INT4), gen_kwargs: (None), limit: 200.0, num_fewshot: None, batch_size: 32

lm_eval --model hf     --model_args pretrained=namgyu-youn/Qwen3-8B     --tasks hellaswag     --device cuda:0     --batch_size 1 --limit 200

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
hellaswag	1	none	0	acc	↑	0.495	±	0.0354
		none	0	acc_norm	↑	0.610	±	0.0346

Paper: TorchAO: PyTorch-Native Training-to-Serving Model Optimization

The model's quantization is powered by TorchAO, a framework presented in the paper TorchAO: PyTorch-Native Training-to-Serving Model Optimization.

Abstract: We present TorchAO, a PyTorch-native model optimization framework leveraging quantization and sparsity to provide an end-to-end, training-to-serving workflow for AI models. TorchAO supports a variety of popular model optimization techniques, including FP8 quantized training, quantization-aware training (QAT), post-training quantization (PTQ), and 2:4 sparsity, and leverages a novel tensor subclass abstraction to represent a variety of widely-used, backend agnostic low precision data types, including INT4, INT8, FP8, MXFP4, MXFP6, and MXFP8. TorchAO integrates closely with the broader ecosystem at each step of the model optimization pipeline, from pre-training (TorchTitan) to fine-tuning (TorchTune, Axolotl) to serving (HuggingFace, vLLM, SGLang, ExecuTorch), connecting an otherwise fragmented space in a single, unified workflow. TorchAO has enabled recent launches of the quantized Llama 3.2 1B/3B and LlamaGuard3-8B models and is open-source at this https URL .

Resources

• Official TorchAO GitHub Repository: https://github.com/pytorch/ao
• TorchAO Documentation: https://docs.pytorch.org/ao/stable/index.html

Disclaimer

PyTorch has not performed safety evaluations or red teamed the quantized models. Performance characteristics, outputs, and behaviors may differ from the original models. Users are solely responsible for selecting appropriate use cases, evaluating and mitigating for accuracy, safety, and fairness, ensuring security, and complying with all applicable laws and regulations.

Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the licenses the models are released under, including any limitations of liability or disclaimers of warranties provided therein.