Testing the speed and size of container image builds.
Look at this summary !
fork this project, modify .github/workflows/main.yml.Of course you can also run it locally.
$ ./build.sh ${DOCKER_FILE_DIR} ${OUTPUT_CSV_PATH} ${RUN_TIMES} ${DIND_IMAGE} ${BUILDER}
# examples:
./build.sh dockerfiles/pytorchcuda out.csv 2 docker:18.09-dind buildkit
./build.sh dockerfiles/pytorchcuda out.csv 2 docker:20.10-dind docker
- docker:18.09-dind (Sep 5,2019)
- docker:19.03-dind (Aug 6,2021)
- docker:20.10-dind (Jun 8,2022)
- buildkit:latest (Now)
- (c and go build 6.7MB)dockerfiles/01
- (pytorch for deep learning 2.07GB) dockerfiles/pytorchnocuda
- (pytorch for deep learning with nvidia-cuda 6.06GB) dockerfiles/pytorchcuda
Core code and processes from https://github.com/AkihiroSuda/buildbench
Many thanks @AkihiroSuda
This code also follows the Apache-2.0 license
.