code_review.txt

set detect.py as the entrancy of reviewing, because dataset is not necessary.

models.yolo,Model.__init__:
1.model cfg is in *.yaml, and it's supposed to call ``check_yaml`` before using it.
2.call ``print_args`` and ``select_device`` 
3.call ``parse_model``
	adventage:
		can gain the width and depth of layers
		print the archieve as
4.check_anchor_order, initialize_weight, initialize_biases may be important
5.print info of model layers, num_params and FLOPs

tips:
	attribute of Path(f): suffix, parent(s), resolve, stem
	
	when run as entrance, such as python path/to/file
		__name__ == '__main__', __file__ == path/to/file
	when import as a module
		__name__ == 'pkg.[subpkg.]module', __file__ == global path of path/to/file 


check_file in utils.general:
	call ``check_suffix`` first
	if input is a file, return the file(str)
	if input is a url