Kudos to both teams PyTorch & Qualcomm
But, I regret to communicate that I prefer to stay with my previous pipeline when porting stuff onto Qualcomm devices (S8G2/3/4, RB5, etc.) in the time being.
After some hours of playing with ExecuTorch, I feel that there are still rough edges.
Nevertheless, I encourage anyone who has never used QNN (or SNPE) to go for ExecuTorch.
Indeed, I’m very much looking forward for the next releases after things are ironed out properly.
Hopefully then, the PyTorch community offers as much as Tensorflow:TFLite.
Relevant links & notes for anyone interested in Qualcomm devices after you’re done with the basic tutorial.