pytorch == 1.3.0
onnx == 1.6.0
caffe2
bash run.sh
There are two hyper-params in caffe_run.py
- batch_size: the batch_size for each iteration. (Default: 1)
- num_of_iterations: how many iterations we need to run to test the speed. (Default: 1000)
I did not change the inference device.(Default: CPU)