为什么tensorboard不能显示所有指标? [英] Why does tensorboard not show all metrics?

问题描述

我正在使用Ray Tune进行超参数优化和记录.Ray Tune成功记录了我的标量值并将其写入Tensorboard日志.

I'm using Ray Tune for hyperparameter optimization and logging. Ray Tune successfully logs my scalar values and writes them to the Tensorboard log.

这些值确实显示在Tensorboards的"SCALARS"部分中，但是在"HPARAMS"部分中，只有 test_acc 显示为度量.

The values do show up in Tensorboards 'SCALARS' section, but in the 'HPARAMS' section, only test_acc shows up as a metric.

这是我记录值的方式:

track.log(pgd_loss=pgd_loss, pgd_acc=pgd_acc, test_acc=test_acc, test_loss=test_loss)

知道可能是什么问题吗?

Any idea what the issue could be?

推荐答案

尝试重新启动tensorboard.Tensorboard似乎在可靠地检测"HPARAMS"部分的新标量值方面存在问题.

Try to restart tensorboard. Tensorboard seems to have an issue reliably detecting new scalar values for the 'HPARAMS' section.

这篇关于为什么tensorboard不能显示所有指标?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！