Classification missing validation accuracy in TensorBoard

[corentin87]

• Network Type: CSPDarknet53 classifier
• TLT Version: 4.0.1

Hi,

I am training a classifier with TAO and the log outputs the following values:
acc, loss, lr, val_acc, val_loss. But when I am visualising using TensorBoard, I am missing the validation accuracy. The training accuracy is present but is kind of useless when watching for overfitting.

Is it an omission from the TAO team? It doesn’t seem like I can do anything to add it.

Thanks for the help

[Morganh]

Could you share the training spec file?

[corentin87]

here is the spec file:
conf_cspdarknet53.yml (1.7 KB)

[Morganh]

Thanks for the info. I can reproduce and will fix it in next release.

[corentin87]

Thanks, when is the next release planned?

[Morganh]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

It is about 2 months.