Install CK

Guide

python3 -m pip install ck

Pull CK repo with MLPerf Docker images

ck pull repo:mlcommons@ck-mlops

Update CK and repos

If you have CK and CK repos already installed, you can updated them as follows:

python3 -m pip install ck -U
ck pull all

MLPerf image classification with TVM

Build Docker image

ck build docker:ck-mlperf-inference-dev-image-classification-onnx-tvm

See available Docker images here.

Note that MLPerf benchmark fails on CentOS-8.

Run Docker image in benchmarking mode

ck run docker:ck-mlperf-inference-dev-image-classification-onnx-tvm

Run Docker image in interactive mode

ck run docker:ck-mlperf-inference-dev-image-classification-onnx-tvm --bash

MLPerf object classification with TVM

Build Docker image

ck build docker:ck-mlperf-inference-dev-object-detection-onnx-tvm

See available Docker images here.

Run Docker image in benchmarking mode

ck run docker:ck-mlperf-inference-dev-object-detection-onnx-tvm

Run Docker image in interactive mode

ck run docker:ck-mlperf-inference-dev-object-detection-onnx-tvm --bash