[Bug] maybe a bug in mmdeploy/codebase/mmdet/models/dense_heads/yolox_head.py

[HeDeYu]

Checklist

• I have searched related issues but cannot get the expected help.
• 2. I have read the FAQ documentation but cannot get the expected help.
• 3. The bug has not been fixed in the latest version.

Describe the bug

yolox转onnx后，动态batch size时，给定输入大于1张图像时会报错。我查看了一下代码发现如下

第97行
max_scores, _ = torch.max(scores, 1)
这里我猜测是需要对任意image（即batch 维度），任意位置（即anchor维度），取所有类别sigmoid后分数最大值（再进一步判断当前图像当前anchor是否有必要输出，即是否>conf threshold），但scores是尺寸为（N，Anchors，K）的tensor，max的dim恐怕是-1（即取categories维度上的max）
即修改为max_scores, _ = torch.max(scores, -1)

然后第99行
scores = scores.where(mask, scores.new_zeros(1))
mask可能需要unsqueeze and repeat一下。
即修改为scores = scores.where(torch.unsqueeze(mask, dim=-1).repeat(1,1,K), scores.new_zeros(1))
K=number of categories (不过可能不用repeat，广播应该是OK的）

另外我对照了mmdet里边对应功能的代码 v3.0.0rc6
mmdet/models/dense_heads/yolox_head.py中的predict_by_feat方法第307行，这里应该是想做同样的事情
max_scores, labels = torch.max(flatten_cls_scores[img_id], 1)
但这里max的dim取1是因为前边flatten_cls_scores[img_id]已经取了img_id，相当于N维度已经不在了，所以dim=1对应categories维度。

Reproduction

详见上节

Environment

。

Error traceback

No response

[RunningLeon]

@HeDeYu Hi, thanks for your feedback. Could you PR us to fix it?

[github-actions]

This issue is marked as stale because it has been marked as invalid or awaiting response for 7 days without any further response. It will be closed in 5 days if the stale label is not removed or if there is no further response.

[github-actions]

This issue is closed because it has been stale for 5 days. Please open a new issue if you have similar issues or you have any new updates now.