in some cases, matmul node will have 1-D cases, which will make onnx_tool.model_profile() crash
as a example, (export an onnx model of YOLOv8-seg).
Then one of matmul nodes might be like below.
This node's secend input have only 1 dim, so we should care nothing about it when counting macs, but program will crash when it try to get the -2 dim of the tensor, this PR try to fix that.
in some cases, matmul node will have 1-D cases, which will make onnx_tool.model_profile() crash as a example, (export an onnx model of YOLOv8-seg). Then one of matmul nodes might be like below. This node's secend input have only 1 dim, so we should care nothing about it when counting macs, but program will crash when it try to get the -2 dim of the tensor, this PR try to fix that.