Sparse Zoo Models benchmarking on cpu

[uberkk]

Hello,

Firstly thank you for the effort.

1. I tried some example models by using benchmark.py. However, I obtain similar results for each yolov3 model. In other words I used three different yolov3 model which are downloaded from SparseZoo, those are pruned, quantized and base. Then I used benchmark.py to see the results and obtain the following results for those models.
   BenchmarkResults:
   items_per_second: 3.8288485463800352
   ms_per_batch: 261.17512559890747
   batch_times_mean: 0.2611751255989075
   batch_times_median: 0.2577577829360962
   batch_times_std: 0.013241334711086931
   End-to-end per image time: 261.17512559890747ms

I set the parameters as follows:
/home/user/Desktop/deepsparse/model/yolov3/model_quantized.onnx --engine onnxruntime --data-path /home/user/Desktop/deneme_test --batch-size 1 --num-iterations 500 --num-warmup-iterations 100 (--quantized-inputs)

(--quantized-inputs) used only for quantized model. What am I missing?

2. How can I test the models and see the annotation for yolov3 or yolov5 with out internet connection or docker. It would be grate if I have a chance.

Thanks!