ModelSpeedupv2 bug lists

I tested nni on several mainstream CNN model zoos. The result seems frustrating.

package	model name	skip_reason
timm	beit	cannot trace (torch._assert)
timm	beitv2	cannot trace (torch._assert)
timm	cait	cannot trace (torch._assert)
timm	coat	cannot trace (torch._assert)
timm	convit	cannot trace (torch._assert)
timm	deit	cannot trace (torch._assert)
timm	dm_nfnet	cannot trace (bool proxy)
timm	eca_nfnet	cannot prune (empty mask)
timm	ese_vovnet	cannot trace (diff # of nodes)
timm	gmixer	cannot trace (torch._assert)
timm	gmlp	cannot trace (torch._assert)
timm	hardcorenas	cannot trace (torch._assert)
timm	mixer	cannot trace (torch._assert)
timm	nf_ecaresnet	cannot prune (empty mask)
timm	nf_regnet	cannot trace (torch._assert)
timm	regnet	cannot trace (torch._assert)
timm	skresnet	cannot trace (diff # of nodes)
timm	swin	cannot trace (torch._assert)
timm	vit	cannot trace (torch._assert)
torchvision	convnext	cannot prune (empty layer)
torchvision	googlenet	output is dict
torchvision	inception	cannot trace
torchvision	shufflenet	cannot prune (endless loop)
torchvision	vit	cannot prune (jit trace error)
torchvision.detection	fasterrcnn	cannot trace (bool proxy)
torchvision.detection	fcos	cannot trace (bool proxy)
torchvision.detection	maskrcnn	cannot trace (bool proxy)
torchvision.detection	retinanet	cannot trace (bool proxy)
torchvision.detection	ssd	cannot trace (bool proxy)
torchvision.detection	keypointrcnn	cannot trace (bool proxy)
torchvision.segmentation	lraspp	cannot prune (fix mask conflict)
torchvision.video	mc3_18	cannot prune (empty mask)
torchvision.video	mvit_v2_s	cannot trace (bool proxy)
torchvision.video	r2plus1d_18	cannot prune (empty mask)
torchvision.video	r3d_18	cannot prune (empty mask)
torchvision.video	s3d	cannot prune (empty mask)