TR0N: Translator Networks for 0-Shot Plug-and-Play Conditional Generation

Zhaoyan Liu, Noël Vouitsis, Satya Krishna Gorti, Jimmy Ba, Gabriel Loaiza-Ganem
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:22092-22112, 2023.

Abstract

We propose TR0N, a highly general framework to turn pre-trained unconditional generative models, such as GANs and VAEs, into conditional models. The conditioning can be highly arbitrary, and requires only a pre-trained auxiliary model. For example, we show how to turn unconditional models into class-conditional ones with the help of a classifier, and also into text-to-image models by leveraging CLIP. TR0N learns a lightweight stochastic mapping which "translates’" between the space of conditions and the latent space of the generative model, in such a way that the generated latent corresponds to a data sample satisfying the desired condition. The translated latent samples are then further improved upon through Langevin dynamics, enabling us to obtain higher-quality data samples. TR0N requires no training data nor fine-tuning, yet can achieve a zero-shot FID of 10.9 on MS-COCO, outperforming competing alternatives not only on this metric, but also in sampling speed – all while retaining a much higher level of generality. Our code is available at https://github.com/layer6ai-labs/tr0n.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-liu23ak, title = {{TR}0{N}: Translator Networks for 0-Shot Plug-and-Play Conditional Generation}, author = {Liu, Zhaoyan and Vouitsis, No\"{e}l and Gorti, Satya Krishna and Ba, Jimmy and Loaiza-Ganem, Gabriel}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {22092--22112}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/liu23ak/liu23ak.pdf}, url = {https://proceedings.mlr.press/v202/liu23ak.html}, abstract = {We propose TR0N, a highly general framework to turn pre-trained unconditional generative models, such as GANs and VAEs, into conditional models. The conditioning can be highly arbitrary, and requires only a pre-trained auxiliary model. For example, we show how to turn unconditional models into class-conditional ones with the help of a classifier, and also into text-to-image models by leveraging CLIP. TR0N learns a lightweight stochastic mapping which "translates’" between the space of conditions and the latent space of the generative model, in such a way that the generated latent corresponds to a data sample satisfying the desired condition. The translated latent samples are then further improved upon through Langevin dynamics, enabling us to obtain higher-quality data samples. TR0N requires no training data nor fine-tuning, yet can achieve a zero-shot FID of 10.9 on MS-COCO, outperforming competing alternatives not only on this metric, but also in sampling speed – all while retaining a much higher level of generality. Our code is available at https://github.com/layer6ai-labs/tr0n.} }
Endnote
%0 Conference Paper %T TR0N: Translator Networks for 0-Shot Plug-and-Play Conditional Generation %A Zhaoyan Liu %A Noël Vouitsis %A Satya Krishna Gorti %A Jimmy Ba %A Gabriel Loaiza-Ganem %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-liu23ak %I PMLR %P 22092--22112 %U https://proceedings.mlr.press/v202/liu23ak.html %V 202 %X We propose TR0N, a highly general framework to turn pre-trained unconditional generative models, such as GANs and VAEs, into conditional models. The conditioning can be highly arbitrary, and requires only a pre-trained auxiliary model. For example, we show how to turn unconditional models into class-conditional ones with the help of a classifier, and also into text-to-image models by leveraging CLIP. TR0N learns a lightweight stochastic mapping which "translates’" between the space of conditions and the latent space of the generative model, in such a way that the generated latent corresponds to a data sample satisfying the desired condition. The translated latent samples are then further improved upon through Langevin dynamics, enabling us to obtain higher-quality data samples. TR0N requires no training data nor fine-tuning, yet can achieve a zero-shot FID of 10.9 on MS-COCO, outperforming competing alternatives not only on this metric, but also in sampling speed – all while retaining a much higher level of generality. Our code is available at https://github.com/layer6ai-labs/tr0n.
APA
Liu, Z., Vouitsis, N., Gorti, S.K., Ba, J. & Loaiza-Ganem, G.. (2023). TR0N: Translator Networks for 0-Shot Plug-and-Play Conditional Generation. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:22092-22112 Available from https://proceedings.mlr.press/v202/liu23ak.html.

Related Material