GPTAQ: Efficient Finetuning-Free Quantization for Asymmetric Calibration

Yuhang Li, Ruokai Yin, Donghyun Lee, Shiting Xiao, Priyadarshini Panda
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:36690-36706, 2025.

Abstract

We introduce GPTAQ, a novel finetuning-free quantization method for compressing large-scale transformer architectures. Unlike the previous GPTQ method, which independently calibrates each layer, we always match the quantized layer’s output to the exact output in the full-precision model, resulting in a scheme that we call asymmetric calibration. Such a scheme can effectively reduce the quantization error accumulated in previous layers. We analyze this problem using optimal brain compression to derive a close-formed solution. The new solution explicitly minimizes the quantization error as well as the accumulated asymmetry error. Furthermore, we utilize various techniques to parallelize the solution calculation, including channel parallelization, neuron decomposition, and Cholesky reformulation for matrix fusion. As a result, GPTAQ is easy to implement, simply using 20 more lines of code than GPTQ but improving its performance under low-bit quantization. Remarkably, on a single GPU, we quantize a 405B language transformer as well as EVA-02—the rank first vision transformer that achieves 90% pretraining Imagenet accuracy. Code is available at Github.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-li25dn, title = {{GPTAQ}: Efficient Finetuning-Free Quantization for Asymmetric Calibration}, author = {Li, Yuhang and Yin, Ruokai and Lee, Donghyun and Xiao, Shiting and Panda, Priyadarshini}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {36690--36706}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/li25dn/li25dn.pdf}, url = {https://proceedings.mlr.press/v267/li25dn.html}, abstract = {We introduce GPTAQ, a novel finetuning-free quantization method for compressing large-scale transformer architectures. Unlike the previous GPTQ method, which independently calibrates each layer, we always match the quantized layer’s output to the exact output in the full-precision model, resulting in a scheme that we call asymmetric calibration. Such a scheme can effectively reduce the quantization error accumulated in previous layers. We analyze this problem using optimal brain compression to derive a close-formed solution. The new solution explicitly minimizes the quantization error as well as the accumulated asymmetry error. Furthermore, we utilize various techniques to parallelize the solution calculation, including channel parallelization, neuron decomposition, and Cholesky reformulation for matrix fusion. As a result, GPTAQ is easy to implement, simply using 20 more lines of code than GPTQ but improving its performance under low-bit quantization. Remarkably, on a single GPU, we quantize a 405B language transformer as well as EVA-02—the rank first vision transformer that achieves 90% pretraining Imagenet accuracy. Code is available at Github.} }
Endnote
%0 Conference Paper %T GPTAQ: Efficient Finetuning-Free Quantization for Asymmetric Calibration %A Yuhang Li %A Ruokai Yin %A Donghyun Lee %A Shiting Xiao %A Priyadarshini Panda %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-li25dn %I PMLR %P 36690--36706 %U https://proceedings.mlr.press/v267/li25dn.html %V 267 %X We introduce GPTAQ, a novel finetuning-free quantization method for compressing large-scale transformer architectures. Unlike the previous GPTQ method, which independently calibrates each layer, we always match the quantized layer’s output to the exact output in the full-precision model, resulting in a scheme that we call asymmetric calibration. Such a scheme can effectively reduce the quantization error accumulated in previous layers. We analyze this problem using optimal brain compression to derive a close-formed solution. The new solution explicitly minimizes the quantization error as well as the accumulated asymmetry error. Furthermore, we utilize various techniques to parallelize the solution calculation, including channel parallelization, neuron decomposition, and Cholesky reformulation for matrix fusion. As a result, GPTAQ is easy to implement, simply using 20 more lines of code than GPTQ but improving its performance under low-bit quantization. Remarkably, on a single GPU, we quantize a 405B language transformer as well as EVA-02—the rank first vision transformer that achieves 90% pretraining Imagenet accuracy. Code is available at Github.
APA
Li, Y., Yin, R., Lee, D., Xiao, S. & Panda, P.. (2025). GPTAQ: Efficient Finetuning-Free Quantization for Asymmetric Calibration. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:36690-36706 Available from https://proceedings.mlr.press/v267/li25dn.html.

Related Material