/benchmarking_edge_tpu

Running benchmark tests on Coral Edge TPU

Primary LanguageJupyter Notebook

benchmarking_edge_tpu

  1. Make sure you have to folders with the co-compiled models and the according inference_reference.csv file within this repo.
  2. Simply run bash run_tests.sh i j where i and j are the different model folders you want to test, so the different combinations, e.g. bash run_tests.sh 0 5 would run the inference test for co-compiled models in folder models_0 up to folder models_5. It is inclusive, so folder 5 will also be tested.

NOTE: After running only inception_v3, the benchmark for this model is about 51 ms.