dl-infer-perf

A perf analysis of deep learning inference performance over pytorch/tensorflow and TensorRT/XLA/TVM.

Environments

TVM

docker: nvidia/cuda:11.1.1-devel-ubuntu18.0

compile tvm with llvm (clang+llvm-11.0.1-x86_64-linux-gnu-ubuntu-16.04)

XLA

docker: nvcr.io/nvidia/tensorflow:20.07-tf2-py3

TensorRT

docker: nvcr.io/nvidia/tensorrt:19.09-py3

Candidates

model

• vgg16
• mobilenet
• resnet50
• inception

optimizer

• TVM
• XLA
• TensorRT

front-end(dl framework)

• ONNX
• Pytorch
• Tensorflow

Usage

Run per optimizer&frontend

usage: python3 infer_perf/<script.py> <model_name> <options>

i.e. run tf resnet50 with xla enabled
python3 infer_perf/to_xla resnet50 --xla

Run varios jobs together

usage: python3 executor <job.json> --report <outout.csv>

i.e.
python executor torch2tvm.json  --report result.csv

Run with benchmark server

usage: python3 executor <job.json> --server <bm_server>

Code Format

yapf infer_perf/*.py -i --style yapf.style