How to show the feature vectors while inference image?

[ryan994]

Hello,
Due to some researches and curious, I tried to print out some feature vectors during inferencing image.
I only find this command when model inferences image:
scores = model(return_loss=False, **data),
which returns a list :
[array([1.2522009e-02, 9.9847050e-08, 9.8747790e-01], dtype=float32)].
what if I want to see the feature vectors of fc layer whild model inferences an image ,how should I do? Not weights in pretrain model file .pth.

Thanks

[mzr1996]

I think you need to use breakpoint or add log output where you want to output feature vector.
For example, if you use LinearClsHead, you need to add it before final fc layer.
https://github.com/open-mmlab/mmclassification/blob/ace09510f700d3f8583e19c993113491cfd10ce4/mmcls/models/heads/linear_head.py#L37-L39

[mzr1996]

Closed by #593