Yolov5_V7.0 while export dynamic onnx, netron layer error

[fenghuoxiguozu]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

use export code:
python export.py --weights yolov5m.pt --include onnx --dynamic --simplify

then i put it into netron, find last Detect layer error

how to solve it

Additional

No response

[UltralyticsAssistant]

👋 Hello @fenghuoxiguozu, thank you for reaching out about YOLOv5 🚀!

It seems like you're encountering an issue with exporting to ONNX with dynamic axes. To help you better, we suggest providing a minimum reproducible example to identify the problem more precisely. 🐛

In the meantime, please ensure that you've followed the installation requirements and that your setup meets the conditions specified in our Requirements📋. Here's a quick start guide:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

Ensure you're using a verified environment. YOLOv5 runs smoothly in several setups, including:

• Notebooks with free GPU:

For a dynamic ONNX export, make sure you've installed all dependencies, including the latest PyTorch and ONNX.

Status

If this badge is green, all CI tests are passing, ensuring reliability.

🔍 Please remember, this is an automated response, but don't worry—an Ultralytics engineer will assist you soon. In the meantime, explore our YOLOv8 for advanced capabilities 🚀!

Looking forward to your minimum reproducible example or any additional details to help resolve your query more effectively.

[fenghuoxiguozu]

class Detect(nn.Module):
# YOLOv5 Detect head for detection models
.....

def forward(self, x):   # [(4,256,80,80),(4,512,40,40),(4,1024,40,40)]
    z = []  # inference output
    for i in range(self.nl):
        x[i] = self.m[i](x[i])  # conv
        bs, _, ny, nx = x[i].shape#map(int, x[i].shape ) # x[i].shape  # x(bs,255,20,20) to x(bs,3,20,20,85)
        x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
        
        # x:[(1,75,32,32),(1,75,16,16),(1,75,8,8)]  =>  [(1,3,32,32,25),(1,3,16,16,25),(1,3,8,8,25)]
        if not self.training:  # inference
            if self.dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:
                self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i)

            if isinstance(self, Segment):  # (boxes + masks)
                xy, wh, conf, mask = x[i].split((2, 2, self.nc + 1, self.no - self.nc - 5), 4)
                xy = (xy.sigmoid() * 2 + self.grid[i]) * self.stride[i]  # xy
                wh = (wh.sigmoid() * 2) ** 2 * self.anchor_grid[i]  # wh
                y = torch.cat((xy, wh, conf.sigmoid(), mask), 4)
            else:  # Detect (boxes only)
                xy, wh, conf = x[i].sigmoid().split((2, 2, self.nc + 1), 4)
                xy = (xy * 2 + self.grid[i]) * self.stride[i]  # xy
                wh = (wh * 2) ** 2 * self.anchor_grid[i]  # wh
                y = torch.cat((xy, wh, conf), 4)
            z.append(y.view(bs, self.na * nx * ny, self.no))

    return x if self.training else (torch.cat(z, 1),) if self.export else (torch.cat(z, 1), x)

[pderrenger]

@fenghuoxiguozu thank you for sharing your code snippet. If you're experiencing issues, please ensure you're using the latest version of YOLOv5. If the problem persists, feel free to provide more details about the specific error or behavior you're encountering.