run_efficientnet.sh

#!/bin/bash
# RUN: %s %t.1 %t.2

model_name="efficientnet"
model_file="$MODELS_ROOT/vision/classification/efficientnet/$model_name-lite4-11.onnx"
image_dir="$MODELS_ROOT/vision/test_images"
if [[ $# != 0 ]];then
  export TEST_TEMP_DIR=`dirname $1`
fi

curr_dir=`dirname $0`

# check if GPU is enabled or not
if [[ $TEST_WITH_GPU -eq 1 ]]; then
  echo "======== Testing with ODLA TensorRT ========"
  for i in 1
  do 
  python3 $curr_dir/../../invoke_halo.py --batch_size $i --model $model_file \
          --label-file $curr_dir/../1000_labels.txt --image-dir $image_dir \
          --odla tensorrt --convert-layout-to=nhwc | tee $1
  done
# RUN: FileCheck --input-file %t.1 %s
fi

# Using HALO to compile and run inference with ODLA XNNPACK
echo "======== Testing with ODLA XNNPACK (NHWC) ========"
python3 $curr_dir/../../invoke_halo.py --model $model_file \
        --label-file $curr_dir/../1000_labels.txt --image-dir $image_dir \
        --odla xnnpack --convert-layout-to=nhwc | tee $2
# RUN: FileCheck --input-file %t.2 %s

# CHECK: dog.jpg ==> "Samoyed, Samoyede",
# CHECK-NEXT: food.jpg ==> "ice cream, icecream",
# CHECK-NEXT: plane.jpg ==> "liner, ocean liner",
# CHECK-NEXT: sport.jpg ==> "ski",