Add reparameterization for yolov7-tiny

[lweicker]

No description provided.

[dsbyprateekg]

I tried your code on the custom yolov7-tiny model.
And its size is reduced to 11.8 MB only from 11.9 MB.

[lweicker]

@dsbyprateekg Once reparameterization is done, you can convert directly the model to engine file, for example in int8, and the output file will be about 8.3MB.

[dsbyprateekg]

@dsbyprateekg Once reparameterization is done, you can convert directly the model to engine file, for example in int8, and the output file will be about 8.3MB.

any link or tutorial for doing this conversion?

[lweicker]

@dsbyprateekg sure, you can check: https://github.com/wang-xinyu/tensorrtx/tree/master/yolov7