Releases: mlrun/mlrun
v0.5.1-rc2
v0.5.1-rc1
Features / Enhancements
- Datasets: Dataset fixes & enhancment, #364, @yaronh
- UI: Features & enhancment
Bug fixes
- Logging: Fixed logger to consider log level correctly, #362, @Hedingber
- CI: Fix UI release branch source, #365, @orkiguazio
- Build: Fix jupyter image, #369, @Hedingber
- Build: Requirements fixes, #373, @Hedingber
- Cleanup: Condition initializing periodic cleanup on whether running in k8s cluster, #367, @Hedingber
- Cleanup: Cleanup error logs fixes, #371, @Hedingber
- Packaging: Fix extra requirements definition, #368, @Hedingber
- Jobs: Enrich owner on job submission, #366, @Hedingber
- UI: Bug fixes
Pull requests:
37f4f85 Bump version to 0.5.1-rc1 (#374)
54f9cf1 Requirements fixes (#373)
9da93d9 Fix jupyter image (#369)
26db92f Lint to exclude git directory (#372)
9bcffd3 Cleanup error logs fixes (#371)
3548f2a Remove latest docker tag references in docs (#370)
14c604f Enrich owner on job submission (#366)
fa173a3 Fix extra requirements definition (#368)
412d705 Condition initializing periodic cleanup on whether running in k8s cluster (#367)
e48bfca Dataset fixes & enhancment (#364)
0c9d8e9 Fix ui release branch source (#365)
782ee6c API docs fixes (#363)
084f0ff Fixing logger to consider log level correctly (#362)
v0.5.0
Features / Enhancements
- Scheduling: Refactor schedules mechanism, #345, @Hedingber
- Datastores: Added support for Azure blobs, #332, @theSaarco
- Artifacts: Extend dataset artifact, #343, @yaronh
- Logging: Support continuous log print in functions with handler, #339, @yaronh
- Logging: Added logger client with kwargs, #297, @liranbg
- Sparkjob: Support spark job types, #350, @urihoenig
- Nuclio runtime: Add nuclio readiness_timeout config, #346, @yaronh
- Cleanup: Clean command improvements, #348, @Hedingber
- Runs: Validate run name regex, #352, @Hedingber
- Monitoring: Add mlrun/scrape_metrics label, #356, @Hedingber
- Build: Dockerfiles fixes and optimizations, #338, @Hedingber
- Code quality: Adding Formatting and Linting using black, #307, @omesser
- CI: Moved to GH actions, #328, @omesser
- CI: Add automation to release mlrun/ui on mlrun/mlrun release, #342, @orkiguazio
Bug fixes
- Nuclio runtime: Fix nuclio remote, secrets, and serialization, #331, @yaronh
- Scheduling: Each job should have different uid, #354, @Hedingber
- Artifacts: Artifacts saved without tag, #351, @Hedingber
- Projects: Fix url parsing to preserve case, #357, @Hedingber
Pull requests:
ac1f096 Bump version to 0.5.0 (#359)
1bd6ae1 Fix mlrun_basics example notebook (#361)
54f2a45 Limit schedules to no more then one job in 10 minutes (#358)
961b211 Fix url parsing to preserve case (#357)
804a573 Add mlrun/scrape_metrics label (#356)
e7ba6f7 Transform from crontab string to schedule trigger in backend (#355)
9af14b1 Validate run name regex (#352)
0f23416 Default artifact tag to latest (#351)
a4b1586 Scheduler fix - each job should have different uid (#354)
326f659 Clean command improvements (#348)
02f9663 Merge pull request #353 from mlrun/nuclio-serve-stream
12a851b when the trigger is not http, pick the first (will be enhanced in future)
637bb78 Add nuclio readiness_timeout config (#346)
4f79606 Use mlrun/mlrun release branch to mlrun/ui release branch (#349)
a8483d1 Support spark job types (#350)
d6210f7 Scheduler fixes (#347)
8d68a53 Refactor schedules mechanism (#345)
da21e32 Add automation to release mlrun/ui on mlrun/mlrun release (#342)
5703014 Extend dataset artifact (#343)
93ff108 Be selective on CI run (#344)
6957930 Handler types like mlrun.DataItem (#341)
28c920f Fix CI (#340)
090c0a4 Added logger client with kwargs (#297)
46f653b Fix dockerfiles and requirements files (#338)
8c127bc Merge branch 'development' of https://github.com/mlrun/mlrun into development
4c18db6 Added binary df for mlutils get_split
2ef9396 support continuous log print in functions with handler (#339)
6bf8780 fix azure blob get, size must be None or >0 (#337)
fc35c1d Fix unstable build (#336)
88771ff fix azure blob overwrite and remote nuclio deploy_step (#334)
c3b43b1 Fix unstable build (#333)
5e160c8 Added support for Azure blobs as an MLRun data store (#332)
380fc58 Fix nuclio remote, add secrets, and serialization (#331)
a74906f Make project's owner field optional (#330)
f357e32 Sparkjob improvements (#329)
b698621 CI - Move to GH actions (#328)
e21adb4 Adding Formatting and Linting using black (#307)
v0.4.10
Features / Enhancements
- Logging: Support setting log level per run. #321, @yaronh
- Monitoring: Add job name label to k8s resources. #324, @Hedingber
- CI: Log docker build commands output. #319, @orkiguazio
- Testing: Add more validations in integration tests. #316, @Hedingber
Bug fixes
- Schedules: Fix scheduled job submission bug. #317, @Hedingber
- Remote work: Fix remote detection. #325, @Hedingber
- Python SDK: Align api client's (httpdb) store_function to return hash key. #323, @Hedingber
- Cleanup: Don't fail cleanup on one resource deletion failure. #322, @Hedingber
v0.4.9
Major features:
- Automatic resources cleanup
- New helper function -
set_environment
to simplify setting environment to work remotely - Simpler process to upgrade mlrun on iguazio systems
- Bunch of bug fixes and docs improvements
Pull requests:
a527d19 Change makefile defaults (#315)
7bd1393 Add environment setup helper function (#312)
194991a Fix default logs path (#314)
ce16f4e Trigger runtimes cleanup periodically (#302)
52ca2eb Default Nuclio function min replicas to 1 (#311)
a168473 Bump version to 0.4.9 (#310)
aa26261 Initial Model Management and Serving doc (#305)
c2e0be1 prioritize hash key over tag (#308)
0d88eb0 Change SQLDB to be the default DB (#309)
261a0a0 fix: mlrun-models-gpu-legacy push correct name (#306)
v0.4.8
Major features:
- new backend based on FastAPI and SQL
- add pipeline watch option and git/slack notifications, gitops
- enhance hyper-params: add list, random, and custom iterations from file or inline
- context.get_child(), context.commit_children() to manage child context and iteration
- load function as an inline python module (function_to_module())
- Enhance log_model/get_model(), add update_model() functions
- simplify and improve remote access and job/workflow submission
- minimal env var settings, load settings from server, simpler auth
- simplify configuration and reduce boilerplate code
- e.g. auto_mount, extend code_to_function
- improve documentation, complete readthedocs guide
- more exception and user error handling
- enhance mlutils lib with pre-baked model evaluation and plots
- add support for mpi (horovod) operator v1
- cli/api for monitoring and delete of function resources (runtimes)
- enhance CI/CD: auto test notebooks, work with remote k8s, jenkins, ..
- a bunch of bug fixes and api enhancements
Pull requests:
e5c55c6 minor-048-fixes (#298)
3eedccc MLRun Docs update (#294)
f1ae9df fix-env-issues (#295)
fceba8a update-mlutils (#292)
164184a Fixing mpijob crd resolve from remote (#290)
13638b4 load-mod-and-fixes (#289)
9116425 Adding runtimes handlers (#285)
7da2b2f Run enhancments (#288)
0757dda Add jenkinsfile (#286)
e27a657 fix get pipelines (return valid dict) (#284)
9dd963e Get pipeline changes (not verbose on wait + refactor) (#281)
54305ec Support for mpijob v1 (#275)
6f1117e fix bug a missing/wrong model (#282)
f74e5e1 Makefile improvements (#276)
fd47502 Add GET pipelines endpoint (#280)
95c95d3 Change path operations with mix os sync and async to be async def and (#279)
9301477 Fixing labels query param typo (#278)
ef9f2f Fix function tags handling (#271)
0d69e4e bring back plot_importance for compatibility (#274)
d1a9d36 Cleanup httpdb test leftovers (#272)
b2334e4 Improve hyperparam + fixes (#273)
ac0dde5 ML images - moving packages to requirements files (#269)
4409614 PHONY targets should be separated by space not comma (#270)
1150218 add verbose print and fix v3io stream (#268)
c015be3 Adding dask notebook to integration test (#267)
ab5d9c5 Scripts -> makefile (#265)
00cd1df Adding manual integration tests (#266)
301862b Refactor httpd to organized app using FastAPI (#255)
8d90457 Mlutils 1 (#264)
1ca2e37 Gitops and remote access improvements (#263)
5f9de0a fix build cli
057ff63 update jupy docker
v0.4.7
Project:
- embed workflow in project ymal
- load project from yaml (in addition to repo)
- support params and secrets in project cli
KFP:
- List workflows API
- TTL and cleanup
- pass env, tag, verbose in deploy_op
- use function links vs in YAML (smaller KFP YAML)
Data
- DataItems: add .as_df() and .local() methods
- support dir logging and .listdir() for local/s3/v3io
- refactor datastore as a separate module
Models and serving:
- add base serving class
- add context.log_model()
- extend ModelArtifact class to support many new attributes and options
- get_model(), generic mechanism to read models + its files and metadata
- inference stream support (record all inference inputs, outputs, metrics, and metadata to a stream)
Dockerfiles
- enhance, doc and centralize containers build (see /dockerfiles)
Other
- Handle credentials/identity forwarding for secure UI & data access
- Bug fixes
v0.4.6
New features and enhancements:
- Support function marketplace (
hub://
) - Enhanced projects support
- add mlutils for auto ML + pre-baked ml images
- simple feature/artifact store (use
store://
paths) - extend context APIs (add log_dataset, etc.)
- improved UI support
- add robustness, HTTP retry, etc.
- various fixes
v0.4.5
v0.4.4
new features and enhancements:
- projects, see new-project and load-project examples
- local_run() call to run functions, notebooks, .. locally
- run and get pipeline options for submitting pipelines remotely
- new UI
- support image pull secrets
- function documentation
- new REST APIs to support UI
- improved/simplify APIs for logging artifacts
- detect version conflicts
- improved dask support
- remove nuclio-sdk dependencies
- add robustness (retry, timeouts, exceptions, ..)
- support numpy types in serialization
- updated/cleaned examples and notebooks
- updated yamls and dockerfiles
- various bug fixes