Releases: mlcommons/mlcube
Releases · mlcommons/mlcube
0.0.10rc2
What's Changed
- user-friendly logging by @hasan7n in #339
- Add mount options from the CLI by @davidjurado in #279
- Fixing CLI commands fail when
mount
option not present. by @sergey-serebryakov in #341 - Add outputfile option to inspect by @hasan7n in #340
- Updating kfp library version to 1.8.22 by @sergey-serebryakov in #349
- Fixing command line interface for the
run
command. by @sergey-serebryakov in #348 - Refactoring code related to MLCube CLI options. by @sergey-serebryakov in #350
- Add CUDA_VISIBLE_DEVICES environment variable when using the --gpus flag by @davidjurado in #345
- Updating GPU (accelerator) support in MLCube. by @sergey-serebryakov in #351
- Initial commit for supporting CLI singularity arguments when input for singularity runner is docker configuration. by @sergey-serebryakov in #352
- Singularity flags bugfix. by @sergey-serebryakov in #353
- Version bump to 0.0.10rc2. by @sergey-serebryakov in #354
New Contributors
Full Changelog: 0.0.10rc1...0.0.10rc2
0.0.10rc1
What's Changed
- Updating MLCube trademark to
R
. by @sergey-serebryakov in #323 - MLC is registered by @TheKanter in #311
- Correct version parsing for singularity-ce. by @sergey-serebryakov in #326
- Correct version comparison in
Client.supports_fakeroot
. by @sergey-serebryakov in #327 - Moving singularity version string parsing into its own method. by @sergey-serebryakov in #328
- Refactoring Platform.get_installed_runner method. by @sergey-serebryakov in #329
- Writing log messages when MLCube fails to validate runner configuration. by @sergey-serebryakov in #330
- Unit test to create Singularity runner config with docker-based MLCube. by @sergey-serebryakov in #331
- Switch to Trusted Publisher method for PyPi releases by @nathanw-mlc in #324
- DockerImage class to work with components of docker image names. by @sergey-serebryakov in #332
- Adding support for
docker:
protocol for DockerImage.from_string mehod. by @sergey-serebryakov in #333 - Parse
short
docker image names correctly. by @sergey-serebryakov in #334 - add podman unit tests by @dfeddema in #312
- Support for arbitrary docker registries. by @sergey-serebryakov in #335
- Support for auth tokens. by @sergey-serebryakov in #336
- Support for docker user name and password environmen variables. by @sergey-serebryakov in #337
- Version bump to 0.0.10rc1. by @sergey-serebryakov in #338
New Contributors
- @TheKanter made their first contribution in #311
Full Changelog: 0.0.10rc0...0.0.10rc1
0.0.10rc0
Release candidate rc0 0.0.10.
- Repository updates
- New internal functionality:
- New method in
Shell
class to run commands and capture their outputs (#313). - Refactoring singularity runner by discarding the
spython
dependency, and adding new class that implements singularity client API (#314). The goal is to better support different methods to run singularity, including support forApptainer
.
- New method in
- New/Improved user-facing functionality:
0.0.9
- Multiple documentation updates (#278, #281, #283, #285).
- Bug fixes (#284, #294, #302, #303).
- Web UI improvements (#273, #274).
- Python runtime and project dependencies
- Support for
~
and environment variables in MLCube parameters (#290). - New CLI arguments for Docker and Singularity runners (
--network
,--security
,--gpus
,--memory
,--cpu
) (#277).
0.0.8
- Support for tasks' custom entry points.
- Unit tests for MLCube core library and docker and singularity runners.
- Bug fixes.
- Support for docker and singularity mount options.
0.0.7
0.0.6
New major features in this release:
- MLCube docker runner can now configure MLCubes using tarballs (these are docker images exported with
docker save
command). - MLCube singularity runner supports run arguments.
- MLCube runners return exit codes when executions fail.
0.0.5
This release introduces a number of bug fixes.
- Updating PyYAML version to
>= 5.4
. - Fixing docker run arguments in Windows OS (mount paths).
- Unifying the use of environment variable pointing to a MLCube system settings file (
MLCUBE_GLOBAL_CONFIG
->MLCUBE_SYSTEM_SETTINGS
).
Also, the MLCube CLI now better handles --task
argument (See PR #224 ).
0.0.4
MLCube config v2
MLCube config v2 is the next iteration of MLCube development that:
- Improves and simplifies MLCube developer experience.
- Improves MLCube user experience.
- Provides new functionality, such as:
- Support for specifying hardware and software requirements in MLCube configuration files.
- Support for multiple platforms of the same type configured for different environments, such as being able to use different GCP accounts to run MLCube in the cloud.