Fatal: Shape and data size don't match

[ChengzhenDu]

Describe the bug 在执行以下命令之后 ncc.exe compile yolo.onnx yolo.kmodel -i onnx -t k210 --dataset train_img

出现报错： Fatal: Shape and data size don't match

注：yolo.onnx文件是由yolov5s模型的pt格式权重文件转换而来。

Environment (please complete the following information):

• OS: WIN10
• nncase version [最新的ci版本]
• DL Framework [e.g. pytorch1.7.0]

Additional context model.zip train_img图片集未知原因不能上传 train_img 中共有181张jpg格式的图片，分辨率为1920*1080

[aclex]

Done a small investigation, according to git bisect it started happening after f339f92af6a2b921f15a57c9b978a8935c9144fb, probably due to changes in src/importer/onnx/ops/constant.cpp.

[zhangyang2057]

1. This is a constant bug, which will be fixed at https://github.com/kendryte/nncase/pull/295
2. You'd better use onnxsim to simplify your onnx model before using ncc.

[zhangyang2057]

Is there any update?

[aclex]

I think, it happened to work for me after the update you've mentioned, so I consider it solved now.

[zhangyang2057]

No response for a long time. Close it.