[Feature] [3.x] May I ask the reason why the color palette value based on the model?

[sanbuphy]

What's the feature?

I see this : https://github.com/open-mmlab/mmdetection/blob/3.x/mmdet/apis/inference.py#L33

        palette (str): Color palette used for visualization. If palette
            is stored in checkpoint, use checkpoint's palette first, otherwise
            use externally passed palette. Currently, supports 'coco', 'voc',
            'citys' and 'random'. Defaults to coco.

If I want to change the config palette I get an error, because the palette use checkpoint's palette first. I don't quite understand the reason for this design , i want to use config palette to override the model palette, is it possible?

Any other context?

No response

[ZwwWayne]

Thanks for your kind suggestion. We will try to fix the orders. Do you think the orders below works for you?

1. config, if it has
2. model state_dict
3. dataset

[sanbuphy]

Thanks for your kind suggestion. We will try to fix the orders. Do you think the orders below works for you?

1. config, if it has
2. model state_dict
3. dataset

Thanks a lot. Yes, I think this order is more intuitive.