Change log#

v1.4.0#

New and updated features#

Functions#

ID

Description

ML-3474

New sub-package in MLRun for packing returning outputs, logging them to MLRun and unpacking inputs, parsing data items to their required type. See packagers, and View in Git.

Projects#

ID

Description

ML-3375

Two new APIs in the MlrunProject object, used to build an image directly through project API, without creating a function and building an image for it: build_config configures the default build for a given project; build_image builds a docker image based on the project configuration. See MlrunProject, Image build configuration, build_image, and View in Git.

ML-4084

New API to run a setup script to enrich a project, when loading the project , and View in Git

Serving#

ID

Description

ML-3654

The error_handler was updated. See Error handling.

UI#

ID

Description

ML-1248

The engine type now displays in the Feature Set Overview tab.

ML-2083

The Run on spot value now displays in the Jobs Overview tab.

ML-3176

The new Passthrough button in the Create Feature Set enables creating a feature set without ingesting its data, previously supported by SDK.

ML-3549

The new Resource monitoring button in the Jobs Details view opens the Grafana dashboard.

ML-3551

Nested workflows (ParallelFor) now fully display in UI.

ML-2922

The Artifacts, Datasets and Models pages have an improved filter. Enhanced look and feel in tables.

Documentation#

ID

Description

ML-3548

Passing parameters between steps using the outputs parameter is now described in Write a pipeline.

ML-3763

The relationship between GPUs and remote functions is now explained in Number of GPUs and Example of Nuclio function.

New documentation pages#

Breaking changes#

ID

Description

ML-3733

mlrun.get_run_db().list_model_endpoints() returns list. Previously, it returned mlrun.api.schemas.model_endpoints.ModelEndpointList.

ML-3773

The aggregation mechanism on Redis databases has improved, but the history of the aggregation (from before the upgrade) is lost, as if there were 0 events during that period.

ML-4053

Pre-v1.4.0: When logging artifacts during a runtime (regular artifacts, not models (ModelArtifact via context.log_model) or datasets (DatasetArtifact via context.log_dataset)), they were strings in the RunObject outputs property. The strings were the target path to the file logged in the artifact. From v1.4.0, they are the store path of the artifact, and not the target path. (They now appear the same as the store paths for logging models and datasets.) This is breaking behavior only if you use the output of the run object as a parameter to another runtime and not as an input. View in Git.

 
 # Set 2 functions:
func1 = project.set_function(...)
func2 = project.set_function(...)

# Run the first function:
run1 = func1.run(...)
# In the function  `func1` we logged a model "my_model" and an artifact "my_artifact"
run1.outputs  
{
    "my_model": "store://...",
    "my_artifact": "store://...",  # Instead of target path: "/User/.../data.csv"
}

# The function `func2` expects a `DataItem` for the logged artifact so passing it through inputs will work as `DataItem` can work with store paths:
run2 = func2.run(..., inputs={"artifact": run1.outputs["my_artifact"]})

# But passing it through a parameter won't work as the string value is now a store path and not a target path:
run2 = func2.run(..., params={"artifact": run1.outputs["my_artifact"]})

Deprecations and future deprecations#

See Deprecations and Future deprecations.

Closed issues#

ID

Description

ML-1787

Optimized distribution of load between chief and workers so that heavy loads do not cause restart of kubelet. View in Git.

ML-2773

Reduced memory footprint for feature vector that joins data from multiple feature sets. View in Git.

ML-3166

New error message when load_project uses an invalid URL source. View in Git.

ML-3420

Fix artifacts corruption due to overflowing size. View in Git.

ML-3443

Spark ingestion engine now supports more than 2 keys in online target. Tech Preview. View in Git.

ML-3470

Changes in secrets are now recorded in the audit log of the platform. View in Git.

ML-3508

Improved description of list_runs. See list_runs View in Git.

ML-3621

clear_context() now does not delete content if the path is relative; and if a subpath exists, only the sub dir is deleted/cleared. View in Git.

ML-3631

MLRun now successfully pulls the source code from gitlab with a personal access token. View in Git.

ML-3652

V3IO_API is now inferred from the DBPATH. View in Git.

ML-3703

project.set_secrets() now throws a file not found exception if the file does not exist. View in Git.

ML-3713

Users can now use pipeline parameters in the spec of jobs created within the workflow py file without causing run failure. View in Git.

ML-3761

**kwargs now forward as expected in MLRun jobs and hyper params. View in Git.

ML-3782

The (incorrect) naming of features causes error when getting the feature vector from the online feature service. The fix is an additional restriction in feature names. See Aggregations View in Git.

ML-3806

Mismatch errors now printed when ingesting from Kafka into offline target. In case of errors (due to type mismatch) no errors are printed.View in Git.

ML-3847

add_code_metadata now prints error messages when working with git View in Git.

ML-3900

Improved error message when ingesting into a feature set (online target) and no features found on retrieval. View in Git.

ML-4129

Errors from BigQuerySource are now forwarded to MLRun. View in Git.

v1.3.3#

Closed issues#

ID

Description

ML-3940

MLRun does not initiate log collection for runs in aborted state. View in Git.

v1.3.2#

Closed issues#

ID

Description

ML-3896

Fixed: MLRun API failed to get pod logs. View in Git.

ML-3865

kubectl now returns logs as expected. View in Git.

ML-3917

Reduced number of logs. View in Git.

ML-3934

Logs are no longer collected for run pods in an unknown state. View in Git.

v1.3.1#

Closed issues#

ID

Description

ML-3764

Fixed the scikit-learn to 1.2 in the tutorial 02-model-training. (Previously pointed to 1.0.) View in Git.

ML-3794

Fixed a Mask detection demo notebook (3-automatic-pipeline.ipynb). View in Git.

ML-3819

Reduce overly-verbose logs on the backend side. View in Git. View in Git.

ML-3823

Optimized /projects endpoint to work faster. View in Git.

Documentation#

New sections describing Git best practices and an example Nuclio function.

v1.3.0#

Client/server matrix, prerequisites, and installing#

The MLRun server is now based on Python 3.9. It’s recommended to move the client to Python 3.9 as well.

MLRun v1.3.0 maintains support for mlrun base images that are based on python 3.7. To differentiate between the images, the images based on python 3.7 have the suffix: -py37. The correct version is automatically chosen for the built-in MLRun images according to the Python version of the MLRun client (for example, a 3.7 Jupyter gets the -py37 images).

MLRun v1.3.x maintains support for mlrun base images that are based on a python 3.7 environment. To differentiate between the images, the images based on python 3.7 have the suffix: -py37. The correct version is automatically chosen for the built-in MLRun images according to the Python version of the MLRun client (for example, a 3.7 Jupyter gets the -py37 images).

For a Python 3.9 environment see Set up a Python 3.9 client environment.

Set up a Python 3.7 client environment (Iguazio versions up to and including v3.5.2)#

Note

There is a known bug with nbformat on the Jupyter version in Iguazio up to and including v3.5.2, which requires upgrading nbformat to 5.7.0. When using an older nbformat, some Jupyter Notebooks do not open.

To install on a Python 3.7 environment (and optionally upgrade to python 3.9 environment):

  1. Configure the Jupyter service with the env variable JUPYTER_PREFER_ENV_PATH=false.

  2. Within the Jupyter service, open a terminal and update conda and pip to have an up-to-date pip resolver.

$CONDA_HOME/bin/conda install -y conda=23.1.0 	
$CONDA_HOME/bin/conda install -y 'pip>=22.0'
$CONDA_HOME/bin/conda install -y nbformat=5.7.0

  1. If you want to upgrade to a Python 3.9 environment, create a new conda env and activate it:

conda create -n python39 python=3.9 ipykernel -y
conda activate python39

  1. Install mlrun:

./align_mlrun.sh

New and updated features#

Feature store#

ID

Description

ML-2592

Offline data can be registered as feature sets (Tech Preview). See Create a feature set without ingesting its data.

ML-2610

Supports SQLSource for batch ingestion (Tech Preview). See SQL data source.

ML-2610

Supports SQLTarget for storey engine (Tech Preview). (Spark is not yet supported.) See SQL target store.

ML-2709

The Spark engine now supports the steps: MapValues, Imputer, OneHotEncoder, DropFeatures; and supports extracting the time parts from the date in the DateExtractor step. See Data transformations.

ML-2802

get_offline_features supports Spark Operator and Remote Spark.

ML-2957

The username and password for the RedisNoSqlTarget are now configured using secrets, as <prefix_>REDIS_USER <prefix_>REDIS_PASSWORD where <prefix> is the optional RedisNoSqlTarget credentials_prefix parameter. See Redis target store.

ML-3008

Supports Spark using Redis as an online KV target, which caused a breaking change.

ML-3373

Supports creating a feature vector over several feature sets with different entities. (Outer joins are Tech Preview.) See Using an offline feature vector. This API will change in a future release, moving the relationship from the feature set to the feature vector.

Logging data#

ID

Description

ML-2845

Logging data using hints. You can now pass data into MLRun and log it using log hints, instead of the decorator. This is the initial change in MLRun to simplify wrapping usable code into MLRun without having to modify it. Future releases will continue this paradigm shift. See more details.

Projects#

ID

Description

ML-3048

When defining a new project from scratch, there is now a default context directory: “./”. This is the directory from which the MLRun client runs, unless otherwise specified.

Serving graphs#

ID

Description

ML-1167

Add support for graphs that split and merge (DAG), including a list of steps for the after argument in the add_step() method. See Graph that splits and rejoins.

ML-2507

Supports configuring of consumer group name for steps following QueueSteps. See Queue (streaming).

Storey#

ID

Description

ML-2502

The event time in storey events is now taken from the timestamp_key. If the timestamp_key is not defined for the event, then the time is taken from the processing-time metadata. View in Git, and in Storey git.

UI#

ID

Description

ML-1186

The new Projects home page provides easy and intuitive access to the full project lifecycle in three phases, with links to the relevant wizards under each phase heading: ingesting and processing data, developing and training a model, deploying and monitoring the project.

mlrun-project-homepage

NA

UI change log in GitHub

APIs#

ID

Description

ML-3104

These APIs now only return reasons in kwargs: log_and_raise, generic_error_handler, http_status_error_handler.

ML-3204

New API set_image_pull_configuration that modifies func.spec.image_pull_secret and func.spec.image_pull_policy, instead of directly accessing these values through the spec.

Documentation#

Improvements to Set up your environment.

Infrastructure improvements#

ID

Description

ML-2609

MLRun server is based on Python 3.9.

ML-2732

The new log collection service improves the performance and reduces heavy IO operations from the API container. The new MLRun log collector service is a gRPC server, which runs as sidecar in the mlrun-api pod (chief and worker). The service is responsible for collecting logs from run pods, writing to persisted files, and reading them on request. The new service is transparent to the end-user: there are no UI or API changes.

Breaking changes#

  • The behavior of ingest with aggregation changed in v1.3.0 (storey, spark, pandas engines). Now, when you ingest a “timestamp” column, it returns
    <class 'pandas._libs.tslibs.timestamps.Timestamp'>.
    Previously, it returned <class 'str'>

  • Any target data that was saved using Redis as an online target with storey engine (RedisNoSql target, introduced in 1.2.1) is not accessible after upgrading to v1.3. (Data ingested subsequent to the upgrade is unaffacted.)

Deprecated and removed APIs#

Starting with v1.3.0, and continuing in subsequent releases, obsolete functions are getting removed from the code. See Deprecations.

Closed issues#

ID

Description

ML-2421

Artifacts logged via SDK with “/” in the name can now be viewed in the UI. View in Git.

ML-2534

Jobs and Workflows pages now display the tag of the executed job (as defined in the API). View in Git.

ML-2810

Fixed the Dask Worker Memory Limit Argument. View in Git.

ML-2896

add_aggregation over Spark fails with AttributeError for sqr and stdvar. View in Git.

ML-3104

Add support for project default image. View in Git.

ML-3119

Fix: MPI job run status resolution considering all workers. View in Git.

ML-3283

project.list_models() did not function as expected for tags and labels. The list_artifacts method now accept a dictionary, and docstrings were added for httpdb and for MLRunProject methods: both list_artifacts and list_models. View in Git.

ML-3286

Fix: Project page displayed an empty list after an upgrade View in Git.

ML-3316

Users with developer and data permissions can now add members to projects they created. (Previously appeared successful in the UI but users were not added). View in Git.

ML-3365 / 3349

Fix: UI Projects’ metrics show N/A for all projects when ml-pipeline is down. View in Git.

ML-3378

Aggregation over a fixed-window that starts at or near the epoch now functions as expected. View in Git.

ML-3380

Documentation: added details on aggregation in windows.

ML-3389

Hyperparams run does not present artifacts iteration when selector is not defined. View in Git.

ML-3424

Documentation: new matrix of which engines support which sources/targets. View in Git.

ML-3575

project.run_function() now uses the argument artifact_path (previously used the project’s configured artifact_path instead). View in Git.

ML-3403

Error on Spark ingestion with offline target without defined path (error: NoneType object has no attribute startswith). Fix: default path defined. View in Git.

ML-3446

Fix: Failed MLRun Nuclio deploy needs better error messages. View in Git.

ML-3482

Fixed model-monitoring incompatibility issue with mlrun client running v1.1.x and a server running v1.2.x. View in Git.

v1.2.3#

Closed issues#

ID

Description

ML-3287

UI now resets the cache upon MLRun upgrades, and the Projects page displays correctly. View in Git.

ML-3801

Optimized /projects endpoint to work faster View in Git.

ML-3819

Reduce overly-verbose logs on the backend side. View in Git.

v1.2.2#

Closed issues#

ID

Description

ML-3797, ML-3798

Fixed presenting and serving large-sized projects. View in Git.

v1.2.1#

New and updated features#

Feature store#

  • Supports ingesting Avro-encoded Kafka records. View in Git.

Third party integrations#

Closed issues#

  • Fix: the Projects | Jobs | Monitor Workflows view is now accurate when filtering for > 1 hour. View in Git.

  • The Kubernetes Pods tab in Monitor Workflows now shows the complete pod details. View in Git.

  • Update the tooltips in Projects | Jobs | Schedule to explain that day 0 (for cron jobs) is Monday, and not Sunday. View in Git.

  • Fix UI crash when selecting All in the Tag dropdown list of the Projects | Feature Store | Feature Vectors tab. View in Git.

  • Fix: now updates next_run_time when skipping scheduling due to concurrent runs. View in Git.

  • When creating a project, the error NotImplementedError was updated to explain that MLRun does not have a DB to connect to. View in Git.

  • When previewing a DirArtifact in the UI, it now returns the requested directory. Previously it was returning the directory list from the root of the container. View in Git.

  • Load source at runtime or build time now fully supports .zip files, which were not fully supported previously.

See more#

v1.2.0#

New and updated features#

Artifacts#

  • Support for artifact tagging:

    • SDK: Add tag_artifacts and delete_artifacts_tags that can be used to modify existing artifacts tags and have more than one version for an artifact.

    • UI: You can add and edit artifact tags in the UI.

    • API: Introduce new endpoints in /projects/<project>/tags.

Auth#

  • Support S3 profile and assume-role when using fsspec.

  • Support GitHub fine grained tokens.

Documentation#

  • Restructured, and added new content.

Feature store#

  • Support Redis as an online feature set for storey engine only. (See Redis target store.)

  • Fully supports ingesting with pandas engine, now equivalent to ingestion with storey engine (TechPreview):

    • Support DataFrame with multi-index.

    • Support mlrun steps when using pandas engine: OneHotEncoder , DateExtractor, MapValue, Imputer and FeatureValidation.

  • Add new step: DropFeature for pandas and storey engines. (TechPreview)

  • Add param query for get_offline_feature for filtering the output.

Frameworks#

  • Add HuggingFaceModelServer to mlrun.frameworks at mlrun.frameworks.huggingface to serve HuggingFace models.

Functions#

  • Add function.with_annotations({"framework":"tensorflow"}) to user-created functions.

  • Add overwrite_build_params to project.build_function() so the user can choose whether or not to keep the build params that were used in previous function builds.

  • deploy_function has a new option of mock deployment that allows running the function locally.

Installation#

  • New option to install google-cloud requirements using mlrun[google-cloud]: when installing MLRun for integration with GCP clients, only compatible packages are installed.

Models#

  • The Labels in the Models > Overview tab can be edited.

Internal#

  • Refactor artifacts endpoints to follow the MLRun convention of /projects/<project>/artifacts/.... (The previous API will be deprecated in a future release.)

  • Add /api/_internal/memory-reports/ endpoints for memory related metrics to better understand the memory consumption of the API.

  • Improve the HTTP retry mechanism.

  • Support a new lightweight mechanism for KFP pods to pull the run state they triggered. Default behavior is legacy, which pulls the logs of the run to figure out the run state. The new behavior can be enabled using a feature flag configured in the API.

Breaking changes#

  • Feature store: Ingestion using pandas now takes the dataframe and creates indices out of the entity column (and removes it as a column in this df). This could cause breakage for existing custom steps when using a pandas engine.

Closed issues#

  • Support logging artifacts larger than 5GB to V3IO. View in Git.

  • Limit KFP to kfp~=1.8.0, <1.8.14 due to non-backwards changes done in 1.8.14 for ParallelFor, which isn’t compatible with the MLRun managed KFP server (1.8.1). View in Git.

  • Add artifact_path enrichment from project artifact_path. Previously, the parameter wasn’t applied to project runs when defining project.artifact_path. View in Git.

  • Align timeouts for requests that are getting re-routed from worker to chief (for projects/background related endpoints). View in Git.

  • Fix legacy artifacts load when loading a project. Fixed corner cases when legacy artifacts were saved to yaml and loaded back into the system using load_project(). View in Git.

  • Fix artifact latest tag enrichment to happen also when user defined a specific tag. View in Git.

  • Fix zip source extraction during function build. View in Git.

  • Fix Docker compose deployment so Nuclio is configured properly with a platformConfig file that sets proper mounts and network configuration for Nuclio functions, meaning that they run in the same network as MLRun. View in Git.

  • Workaround for background tasks getting cancelled prematurely, due to the current FastAPI version that has a bug in the starlette package it uses. The bug caused the task to get cancelled if the client’s HTTP connection was closed before the task was done. View in Git.

  • Fix run fails after deploying function without defined image. View in Git.

  • Fix scheduled jobs failed on GKE with resource quota error. View in Git.

  • Can now delete a model via tag. View in Git.

See more#

v1.1.3#

Closed issues#

  • The CLI supports overwriting the schedule when creating scheduling workflow. View in Git.

  • Slack now notifies when a project fails in load_and_run(). View in Git.

  • Timeout is executed properly when running a pipeline in CLI. View in Git.

  • Uvicorn Keep Alive Timeout (http_connection_timeout_keep_alive) is now configurable, with default=11. This maintains API-client connections. View in Git.

See more#

v1.1.2#

New and updated features#

V3IO

  • v3io-py bumped to 0.5.19.

  • v3io-fs bumped to 0.1.15.

See more#

v1.1.1#

New and updated features#

API#

  • Supports workflow scheduling.

UI#

  • Projects: Supports editing model labels.

See more#

v1.1.0#

New and updated features#

API#

  • MLRun scalability: Workers are used to handle the connection to the MLRun database and can be increased to improve handling of high workloads against the MLRun DB. You can configure the number of workers for an MLRun service, which is applied to the service’s user-created pods. The default is 2.

    • v1.1.0 cannot run on top of 3.0.x.

    • For Iguazio versions prior to v3.5.0, the number of workers is set to 1 by default. To change this number, contact support (helm-chart change required).

    • Multi-instance is not supported for MLrun running on SQLite.

  • Supports pipeline scheduling.

Documentation#

Feature store#

  • Supports S3, Azure, GCS targets when using Spark as an engine for the feature store.

  • Snowflake as datasource has a connector ID: iguazio_platform.

  • You can add a time-based filter condition when running get_offline_feature with a given vector.

Storey#

  • MLRun can write to parquet with flexible schema per batch for ParquetTarget: useful for inconsistent or unknown schema.

UI#

  • The Projects home page now has three tiles, Data, Jobs and Workflows, Deployment, that guide you through key capabilities of Iguazio, and provide quick access to common tasks.

  • The Projects | Jobs | Monitor Jobs tab now displays the Spark UI URL.

  • The information of the Drift Analysis tab is now displayed in the Model Overview.

  • If there is an error, the error messages are now displayed in the Projects | Jobs | Monitor jobs tab.

Workflows#

  • The steps in Workflows are color-coded to identify their status: blue=running; green=completed; red=error.

See more#

v1.0.6#

Closed issues#

  • Import from mlrun fails with “ImportError: cannot import name dataclass_transform”. Workaround for previous releases: Install pip install pydantic==1.9.2 after align_mlrun.sh.

  • MLRun FeatureSet was not enriching with security context when running from the UI. View in Git.

  • MlRun Accesskey presents as cleartext in the mlrun yaml, when the mlrun function is created by feature set request from the UI. View in Git.

See more#

v1.0.5#

Closed issues#

  • MLRun: remove root permissions. View in Git.

  • Users running a pipeline via CLI project run (watch=true) can now set the timeout (previously was 1 hour). View in Git.

  • MLRun: Supports pushing images to ECR. View in Git.

See more#

v1.0.4#

New and updated features#

  • Bump storey to 1.0.6.

  • Add typing-extensions explicitly.

  • Add vulnerability check to CI and fix vulnerabilities.

Closed issues#

  • Limit Azure transitive dependency to avoid new bug. View in Git.

  • Fix GPU image to have new signing keys. View in Git.

  • Spark: Allow mounting v3io on driver but not executors. View in Git.

  • Tests: Send only string headers to align to new requests limitation. View in Git.

See more#

v1.0.3#

New and updated features#

  • Jupyter Image: Relax artifact_path settings and add README notebook. View in Git.

  • Images: Fix security vulnerabilities. View in Git.

Closed issues#

  • API: Fix projects leader to sync enrichment to followers. View in Git.

  • Projects: Fixes and usability improvements for working with archives. View in Git.

See more#

v1.0.2#

New and updated features#

  • Runtimes: Add java options to Spark job parameters. View in Git.

  • Spark: Allow setting executor and driver core parameter in Spark operator. View in Git.

  • API: Block unauthorized paths on files endpoints. View in Git.

  • Documentation: New quick start guide and updated docker install section. View in Git.

Closed issues#

  • Frameworks: Fix to logging the target columns in favor of model monitoring. View in Git.

  • Projects: Fix/support archives with project run/build/deploy methods. View in Git.

  • Runtimes: Fix jobs stuck in non-terminal state after node drain/pre-emption. View in Git.

  • Requirements: Fix ImportError on ingest to Azure. View in Git.

See more#

v1.0.0#

New and updated features#

Feature store#

  • Supports snowflake as a datasource for the feature store.

Graph#

  • A new tab under Projects | Models named Real-time pipelines displays the real time pipeline graph, with a drill-down to view the steps and their details. [Tech Preview]

Projects#

  • Setting owner and members are in a dedicated Project Settings section.

  • The Project Monitoring report has a new tile named Consumer groups (v3io streams) that shows the total number of consumer groups, with drill-down capabilities for more details.

Resource management#

  • Supports preemptible nodes.

  • Supports configuring CPU, GPU, and memory default limits for user jobs.

UI#

  • Supports configuring pod priority.

  • Enhanced masking of sensitive data.

  • The dataset tab is now in the Projects main menu (was previously under the Feature store).

See more#

Open issues#

ID

Description

Workaround

Opened in

ML-1584

Cannot run code_to_function when filename contains special characters

Do not use special characters in filenames

v1.0.0

ML-2199

Spark operator job fails with default requests args.

NA

v1.0.0

ML-2223

Cannot deploy a function when notebook names contain “.” (ModuleNotFoundError)

Do not use “.” in notebook name

v1.0.0

ML-2407

Kafka ingestion service on an empty feature set returns an error.

Ingest a sample of the data manually. This creates the schema for the feature set and then the ingestion service accepts new records.

v1.1.0

ML-2489

Cannot pickle a class inside an mlrun function.

Use cloudpickle instead of pickle

v1.2.0

2621

Running a workflow whose project has init_git=True, results in Project error

Run git config --global --add safe.directory '*' (can substitute specific directory for *).

v1.1.0

ML-3315

Spark ingestion does not support nested aggregations.

NA

v1.2.1

ML-3386

Documentation is missing full details on the feature store sources and targets

NA

v1.2.1

ML-3445

project.deploy_function operation might get stuck when running v1.3.0 demos on an Iguazio platform running v3.2.x.

Replace code: serving_fn = mlrun.new_function("serving", image="python:3.9", kind="serving", requirements=["mlrun[complete]", "scikit-learn~=1.2.0"]) with:
function = mlrun.new_function("serving", image="python:3.9", kind="serving") function.with_commands([ "python -m pip install --upgrade pip", "pip install 'mlrun[complete]' scikit-learn==1.1.2", ])

v1.3.0

ML-3480

Documentation: request details on label parameter of feature set definition

NA

v1.2.1

NA

The feature store does not support schema evolution and does not have schema enforcement.

NA

v1.2.1

ML-3633

Fail to import a context from dict

When loading a context from dict (e.g.: mlrun.MLClientCtx.from_dict(context)), make sure to provide datetime objects and not string. Do this by executing context['status']['start_time'] = parser.parse(context['status']['start_time'])<br> context['status']['last_update'] = parser.parse(context['status']['last_update']) prior to mlrun.MLClientCtx.from_dict(context)

v1.3.0

ML-3640

When running a remote function/workflow, the context global parameter is not automatically injected.

Use get_or_create_ctx

1.3.0

ML-2030

Need means of moving artifacts from test to production Spark

To register artifact between different environments, e.g. dev and prod, upload your artifacts to a remote storage, e.g. S3. You can change the project artifact path using MLRun or MLRun UI. project.artifact_path='s3:<bucket-name/..'

1.0.0

ML-2380

Spark runtime should sustain naive user actions

NA

1.0.4

Limitations#

ID

Description

Workaround

Opened in

ML-2014

Model deployment returns ResourceNotFoundException (Nuclio error that Service is invalid.)

Verify that all metadata.labels values are 63 characters or less. See the Kubernetes limitation.

v1.0.0

ML-3381

Private repo is not supported as a marketplace hub

See Import and run the function from your repo.

v1.2.1

ML-3520

MLRun does not decompress large Kubeflow pipelines

NA

v1.3.0

ML-3824

MLRun supports TensorFlow up to 2.11.

NA

v1.3.1

ML-3731

When trying to identify a failed step in a workflow with mlrun.get_run_db().list_pipelines('project-name'), the returned error is None.

To see the error, use mlrun.db.get_pipelines() instead.

ML-3743

Setting AWS credentials as project secret cause a build failure on EKS configured with ECR.

When using an ECR as the external container registry, make sure that the project secrets AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY have read/write access to ECR, as described in the platform documentation

Deprecations#

In

ID

Description

v1.0.0

NA

MLRun / Nuclio do not support python 3.6.

v1.3.0

NA

Deprecated and removed APIs and Deprecated REST APIs, APIs deprecated in v1.3.0, will be removed in v1.5.0, CLI Deprecated in v1.3.0, will be removed in v1.5.0.

v1.4.0

ML-3547

APIs deprecated in v1.4.0, will be removed from v1.6.0 code and REST APIs deprecated in v1.4.0, will be removed from v1.6.0 code.

Future deprecations#

ID

When

Description

ML-3605

v1.5.0

Model Monitoring: Most of the charts and KPIs in Grafana are now based on the data store target instead of the MLRun API. It is recommended to update the model monitoring dashboards since the old dashboards won’t be supported.

Deprecated APIs and CLI#

APIs deprecated in v1.4.0, will be removed from v1.6.0 code#

These APIs will be removed from the v1.6.0 code. A FutureWarning appears if you try to use them in v1.4.0 and higher.

Deprecated / to be removed

Use instead

MLRunProject.clear_context()

This method deletes all files and clears the context directory or subpath (if defined). This method can produce unexpected outcomes and is not recommended.

MLRunProject object legacy parameters

metadata and spec instead

BaseRuntime.with_commands and KubejobRuntime.build_config ‘verify_base_image’ param

‘prepare_image_for_deploy’

run_local

function.run(local=True)

REST APIs deprecated in v1.4.0, will be removed from v1.6.0 code#

Deprecated

Use instead

POST /artifact/{project}/{uid}/{key:path}

/projects/{project}/artifacts/{uid}/{key:path} instead

GET /projects/{project}/artifact/{key:path}

/projects/{project}/artifacts/{key:path} instead

DELETE /artifact/{project}/{uid}

/projects/{project}/artifacts/{uid} instead

GET /artifacts

/projects/{project}/artifacts instead

DELETE /artifacts

/projects/{project}/artifacts instead

POST /func/{project}/{name}

/projects/{project}/functions/{name} instead

GET /func/{project}/{name}

/projects/{project}/functions/{name} instead

GET /funcs

/projects/{project}/functions instead

APIs deprecated and removed from v1.3.0 code#

These MLRun APIs have been deprecated since at least v1.0.0 and were removed from the code:

Deprecated/removed

Use instead

project.functions

project.get_function, project.set_function, project.list_function

project.artifacts

project.get_artifact, project.set_artifact, project.list_artifact

project.func()

project.get_function()

project.create_vault_secrets()

NA

project.get_vault_secret()

NA

MlrunProjectLegacy class

MlrunProject

Feature-store: usage of state in graph. For example: add_writer_state, and the after_state parameter in _init_ methods.

step

mount_path parameter in mount_v3io()

volume_mounts

NewTask

new_task()

Dask with_limits

with_scheduler_limits / with_worker_limits

Dask with_requests

with_scheduler_requests / with_worker_requests

REST APIs deprecated and removed from v1.3.0 code#

  • pod_status header from response to get_log REST API

  • client-spec from response to health API

  • submit_pipeline_legacy REST API

  • get_pipeline_legacy REST API

  • Five runtime legacy REST APIs, such as: list_runtime_resources_legacy, delete_runtime_object_legacy etc.

  • httpdb runtime-related APIs using the deprecated runtimes REST APIs, for example: delete_runtime_object

APIs deprecated in v1.3.0, will be removed in v1.5.0#

These APIs will be removed from the v1.5.0 code. A FutureWarning appears if you try to use them in v1.3.0 and higher.

Deprecated / to be removed

Use instead

project-related parameters of set_environment. (Global-related parameters will not be deprecated.)

The same parameters in project-related APIs, such as get_or_create_project

KubeResource.gpus

with_limits

Dask gpus

with_scheduler_limits / with_worker_limits

ExecutorTypes

ParallelRunnerModes

Spark runtime gpus

with_driver_limits / with_executor_limits

mount_v3io_legacy (mount_v3io no longer calls it)

mount_v3io

mount_v3io_extended

mount_v3io

LegacyArtifact and all legacy artifact types that inherit from it (LegacyArtifact, LegacyDirArtifact, LegacyLinkArtifact, LegacyPlotArtifact, LegacyChartArtifact, LegacyTableArtifact, LegacyModelArtifact, LegacyDatasetArtifact, LegacyPlotlyArtifact, LegacyBokehArtifact)

Artifact or other artifact classes that inherit from it

init_functions in pipelines

Add the function initialization to the pipeline code instead

The entire mlrun/mlutils library

mlrun.framework

run_pipeline

project.run

CLI Deprecated in v1.3.0, will be removed in v1.5.0#

The --ensure-project flag of the mlrun project CLI command is deprecated and will be removed in v1.5.0.