TensorInferenceBenchmarks

A package for generating benchmark and test reports for TensorInference.jl.

Benchmarks

We use PkgBenchmark.jl to run the benchmark suite defined in TensorInferenceBenchmarks/benchmark/benchmarks.jl. This benchmark uses the networks presented in the UAI 2014 Inference Competition .

To start a benchmark instance, activate the TensorInferenceBenchmarks.jl project and run:

using TensorInferenceBenchmarks
benchmark()

The results of each benchmark instance is stored in a newly created directory inside TensorInferenceBenchmarks/benchmark/benchmark_results/. The name of this new directory is based on the host name, and current date and time.

Test reports

The test reports are generated using TestReports.jl. To generate a test report, activate the TensorInferenceBenchmarks.jl project and run:

using TensorInferenceBenchmarks
test()

The tests that are run are the standard Julia package tests for TensorInferenceBenchmarks.jl stored inside the test directory. Each test report is stored in a newly created directory inside TensorInferenceBenchmarks/test/test_reports/. The name of this new directory is based on the host name, and current date and time.