MLOps integration for MaskRCNN

[masi]

Hello,

I have some issues integrating WandB and Tensorboard into MaskRCNN experiments as the network does not have a “training_config” element in its config file. Is there any other way to integrate an MLOps tool for MaskRCNN so I can track my experiments?

Thank you in advance.

[Morganh]

Please set wandb_config or clearml_config under root level.

[masi]

It does not work. I tried to add wandb_config under the “visualizer” element also on root level (in the MaskRCNN config) and the field name is not recognized. When I try to add a “training_config” element is not recognized either.

This is the error output:

[Morganh]

More, it is needed to use the 4.0.1 version of tao container.
You can pull it from (http://nvcr.io/nvidia/tao/tao-toolkit:4.0.1-tf1.15.5)
Then run “docker run --runtime -it --rm nvcr.io/nvidia/tao/tao-toolkit:4.0.1-tf1.15.5 /bin/bash”

[masi]

Thanks for the help. I was using version 4.0.0 so that explains it.