the latency issues are what usually kill me when trying to run smth heavy on mobile. i tried implementing a transformer-based model once and the
battery drain was insane bc of the constant processing. are you planning to handle the inference on a backend or are you trying to run it locally via smth like
onnxruntime
?