3

I've currently fit a model using sktime (a scikit like library) which provides interfaces to dump and load pickled models. However if I fit in x86 (64 bit OS) and run inference on raspberry pi zero which is armv6 (32 bit OS) the predictions are completely different on the same test set. Digging deeper into the issue I found we have to export in either PMML or ONNX format and then inference using that for cross-architecture operations. The problem is PMML is limited and doesn't support sktime functions and ONNX Runtime Inference is currently broken for raspberry pi zero w (armv6). What are my options?

Rakshith G B
  • 816
  • 2
  • 8
  • 24
  • How did you install sktime on raspberry pi? Could you please try to answer this question? https://stackoverflow.com/questions/76659428/how-to-install-sktime-on-raspberry-pi – Sashi Jul 11 '23 at 07:06

0 Answers0