About BanditPolicy

Question

Hello

After read BanditPolicy doc description, I still had no idea what it is.
Please help to explain more in detail.
Or provide a simple example to illustrate.

I have a request to compare local yolov5 method.
But it seems there is different earlystopping method.
So the training epoch are different.

How to fix azure AutoML yolov5 training epoch?

Thanks!

Answer 1

Hello

Thanks for your prompt reply.

But I just want to focus on BanditPolicy for earlystopping.

"early_termination_policy": BanditPolicy(
    evaluation_interval=2, slack_factor=0.2, delay_evaluation=6

After check azure doc, I still had no idea how to setting these parameters.
Would you help a simple example to illustrate.

Furthermore, how do I set the config that make only train 100 epochs?

Thanks