arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2203.13751cs.LGcs.CV
收藏

高效-VDVAE:少即是多

Efficient-VDVAE: Less is more

Cash App Labs(Cash App实验室)

Louay Hazami, Rayhane Mama, Ragavan Thurairatnam

摘要:

近年来,分层VAE已成为最大似然估计的可靠选择。然而,不稳定性问题和苛刻的计算要求阻碍了该领域的研究进展。我们对Very Deep VAE进行了简单的修改,使其更快地收敛到$2.6\times$,在内存负载中节省$20\times$,并提高训练过程中的稳定性。尽管有这些变化,我们的模型实现了可比或更好的负对数似然性能比当前国家的最先进的模型上的所有$7$常用的图像datasets,我们evaluatedon. We还反对使用5位的基准作为一种方式来衡量分层VAE的性能,由于5位量化所造成的不良偏差的参数。此外,我们的经验表明,大致$3\%$的分层VAE的潜在空间尺寸是足够的编码的图像信息,而不损失性能,打开大门,有效地利用分层VAE的潜在空间在下游任务。我们在https://github.com/Rayhane-mamah/Efficient-VDVAE上发布源代码和模型。

英文摘要:

Hierarchical VAEs have emerged in recent years as a reliable option for maximum likelihood estimation. However, instability issues and demanding computational requirements have hindered research progress in the area. We present simple modifications to the Very Deep VAE to make it converge up to $2.6\times$ faster, save up to $20\times$ in memory load and improve stability during training. Despite these changes, our models achieve comparable or better negative log-likelihood performance than current state-of-the-art models on all $7$ commonly used image datasets we evaluated on. We also make an argument against using 5-bit benchmarks as a way to measure hierarchical VAE's performance due to undesirable biases caused by the 5-bit quantization. Additionally, we empirically demonstrate that roughly $3\%$ of the hierarchical VAE's latent space dimensions is sufficient to encode most of the image information, without loss of performance, opening up the doors to efficiently leverage the hierarchical VAEs' latent space in downstream tasks. We release our source code and models at https://github.com/Rayhane-mamah/Efficient-VDVAE .

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑