README.rst

Overview

docs
tests
package

A small python3 based benchmark tool to benchmark and hence compare Deep Learning Models. This project is currently heavy development. Please expect the usage to change without prior notice.

• Free software: BSD 3-Clause License

Installation

(currently not available)

pip install deeprace

Usage

$ deeprace --help
usage: deeprace [--version] [--help] [--verbose] [--loglevel <level>] <command> [<args>...]

options:
   -h, --help                           Show this help message
   -v, --version                        Print the version of deeprace
   -V, --verbose                        Run in verbose mode
   -L <level> --loglevel=<level>        logging level to use [default: info]

The most commonly used git commands are:
   list      list available models
   train     run training on a given model
   describe  show available parameters for given model

See 'deeprace help <command>' for more information on a specific command.

To run training, do:

$ python3 ./deeprace.py train resnet50v1

usage: deeprace train [options] [--] <models>

options:
    -h, --help                                 print this help message
    -O <mopts> --meta-options=<mopts>          hyper-parameters for training, e.g. batch_size
    -D <dpath> --datapath=<dpath>              path used for temporary storage, e.g. for the input data, checkpoints etc [default: datasets]
    -R <rpath> --resultspath=<rpath>           path to store results or checkpoints [default: deeprace-results]
    -b <backe> --backend=<backe>               which backend to use [default: keras]
    -e <neps> --nepochs=<neps>                 number of epochs to train [default: 0]
    -d <ds> --dataset=<ds>                     the dataset to use (depends on the model of choice) [default: model_default]
    -f <dfrac> --datafraction=<dfrac>          fraction of the dataset to use, helpful for debugging/testing [default: 1.]
    -t <output> --timings=<output>             file to store the individual timings in [default: timings.tsv]
    -s <sep> --separator=<sep>                 seperator for the output data [default:       ]
    -c <cmt> --comment=<cmt>                   comment to add to the measurement

Documentation

(currently not available)

https://deeprace.readthedocs.io/

Development

To run the all tests run:

tox

Note, to combine the coverage data from all the tox environments run:

Windows	set PYTEST_ADDOPTS=--cov-append tox
Other	PYTEST_ADDOPTS=--cov-append tox