Combine multiple pretrained models for ASR

1

hi every one
my question is how to combine output of two pretrained pytorch model like hubert and wav2vec2xlsr model for ASR in arabic language? and then add a CTC layer to the resulted model?
thanks to all