Input arguments and close mosaic option

[Marco-Nguyen]

Hi, I would like to ask 2 questions:

1. I am using Python. Where can I find the list of full input arguments for model.train(), model.val(), model.predict(), and model.export()?
      I have searched in the repo, official docs website but the provided lists do not contain all arguments. For example, device argument can be used in prediction or validation but not listed.

2. What is the purpose of disabling mosaic augmentation for some final epochs? Is there any guidance on how to set the value? and where can I read more on this?

[Laughing-q]

@Marco-Nguyen hi you can find the args here. https://github.com/ultralytics/ultralytics/blob/main/ultralytics/yolo/cfg/default.yaml or in our docs https://docs.ultralytics.com/cfg/.
About the close-mosaic, the idea is from YOLOX and we found it's also benefiting our model. We set close-mosaic=10 as default, I think mostly you don't need to modify it, but if you're using less epochs like 50 epochs or less, then I recommend you modify it to a small number .i.e 5.

[Marco-Nguyen]

@Laughing-q Thanks for the response.

For question 1, as I mentioned, device is not listed neither in the docs nor default.yaml of predict and val. Also in predict, I can pass conf to change the default confidence threshold, and it is not listed as well. Will you update this in docs?

[Laughing-q]

@Marco-Nguyen sure! :)