Microsoft Azure Batch AI Management Client Library for Python
Project description
Microsoft Azure SDK for Python
This package has been deprecated and will no longer be maintained after 10-31-2024. This package will only receive security fixes until 10-31-2024. To receive updates on new features and non-security bug fixes, upgrade to the replacement package, azureml-core. Refer to the migration guide (https://learn.microsoft.com/previous-versions/azure/batch-ai/how-to-migrate) for guidance on upgrading.
Release History
7.0.0 (2024-10-31)
Other Changes
- This package has been deprecated and will no longer be maintained after 10-31-2024. This package will only receive security fixes until 10-31-2024. To receive updates on new features and non-security bug fixes, upgrade to the replacement package, azureml-core. Refer to the migration guide (https://learn.microsoft.com/previous-versions/azure/batch-ai/how-to-migrate) for guidance on upgrading.
7.0.0b1 (2021-06-02)
This is beta preview version. For detailed changelog please refer to equivalent stable version 2.0.0(https://pypi.org/project/azure-mgmt-batchai/2.0.0/)
This version uses a next-generation code generator that introduces important breaking changes, but also important new features (like unified authentication and async programming).
General breaking changes
-
Credential system has been completly revamped:
azure.common.credentials
ormsrestazure.azure_active_directory
instances are no longer supported, use theazure-identity
classes instead: https://pypi.org/project/azure-identity/credentials
parameter has been renamedcredential
-
The
config
attribute no longer exists on a client, configuration should be passed as kwarg. Example:MyClient(credential, subscription_id, enable_logging=True)
. For a complete set of supported options, see the parameters accept in init documentation of azure-core -
You can't import a
version
module anymore, use__version__
instead -
Operations that used to return a
msrest.polling.LROPoller
now returns aazure.core.polling.LROPoller
and are prefixed withbegin_
. -
Exceptions tree have been simplified and most exceptions are now
azure.core.exceptions.HttpResponseError
(CloudError
has been removed). -
Most of the operation kwarg have changed. Some of the most noticeable:
raw
has been removed. Equivalent feature can be found usingcls
, a callback that will give access to internal HTTP response for advanced user- For a complete set of supported options, see the parameters accept in Request documentation of azure-core
General new features
- Type annotations support using
typing
. SDKs are mypy ready. - This client has now stable and official support for async. Check the
aio
namespace of your package to find the async client. - This client now support natively tracing library like OpenCensus or OpenTelemetry. See this tracing quickstart for an overview.
2.0.0 (2018-06-07)
Breaking changes
This version uses 2018-05-01 BatchAI API specification which introduced the following braking changes:
- Clusters, FileServers must be created under a workspace;
- Jobs must be created under an experiment;
- Clusters, FileServers and Jobs do not accept location during creation and belong to the same location as the parent workspace;
- Clusters, FileServers and Jobs do not support tags;
- BatchAIManagementClient.usage renamed to BatchAIManagementClient.usages;
- Job priority changed a type from int to an enum;
- File.is_directory is replaced with File.file_type;
- Job.priority and JobCreateParameters.priority is replaced with scheduling_priority;
- Removed unsupported MountSettings.file_server_type attribute;
- OutputDirectory.type unsupported attribute removed;
- OutputDirectory.create_new attributes removed, BatchAI will always create output directories if they not exist;
- SetupTask.run_elevated attribute removed, the setup task is always executed under root.
Features
- Added support to workspaces to group Clusters, FileServers and Experiments and remove limit on number of allocated resources;
- Added support for experiment to group jobs and remove limit on number of jobs;
- Added support for configuring /dev/shm for jobs which use docker containers;
- Added first class support for generic MPI jobs;
- Added first class support for Horovod jobs.
1.0.1 (2018-04-16)
Bugfixes
- Fix some invalid models in Python 3
- Compatibility of the sdist with wheel 0.31.0
1.0.0 (2018-03-19)
General Breaking changes
This version uses a next-generation code generator that might introduce breaking changes.
-
Model signatures now use only keyword-argument syntax. All positional arguments must be re-written as keyword-arguments. To keep auto-completion in most cases, models are now generated for Python 2 and Python 3. Python 3 uses the "*" syntax for keyword-only arguments.
-
Enum types now use the "str" mixin (class AzureEnum(str, Enum)) to improve the behavior when unrecognized enum values are encountered. While this is not a breaking change, the distinctions are important, and are documented here: https://docs.python.org/3/library/enum.html#others At a glance:
- "is" should not be used at all.
- "format" will return the string value, where "%s" string formatting will return
NameOfEnum.stringvalue
. Format syntax should be prefered.
-
New Long Running Operation:
- Return type changes from
msrestazure.azure_operation.AzureOperationPoller
tomsrest.polling.LROPoller
. External API is the same. - Return type is now always a
msrest.polling.LROPoller
, regardless of the optional parameters used. - The behavior has changed when using
raw=True
. Instead of returning the initial call result asClientRawResponse
, without polling, now this returns an LROPoller. After polling, the final resource will be returned as aClientRawResponse
. - New
polling
parameter. The default behavior isPolling=True
which will poll using ARM algorithm. WhenPolling=False
, the response of the initial call will be returned without polling. polling
parameter accepts instances of subclasses ofmsrest.polling.PollingMethod
.add_done_callback
will no longer raise if called after polling is finished, but will instead execute the callback right away.
- Return type changes from
Features
- added support for job level mounting
- added support for environment variables with secret values
- added support for performance counters reporting in Azure Application Insights
- added support for custom images
- added support for pyTorch deep learning framework
- added API for usage and limits reporting
- added API for listing job files in subdirectories
- now user can choose caching type during NFS creation
- get cluster now reports a path segment generated for storing start task output logs
- get job now reports a path segment generated for job's output directories
- renamed EnvironmentSetting to EnvironmentVariable
0.2.0 (2017-10-05)
- credentials_info property got renamed to credentials.
- removed unused class FileServerStatus and Code enum
- renamed enums for CachingType and VmPriority
- removed 'statuses' attribute on FileServer
0.1.0 (2017-10-03)
- Initial Release
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for azure_mgmt_batchai-7.0.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 9ed3f4117f3a182281b524debb7803aa108f0b0301e908a4cb1ed78c49fc9d19 |
|
MD5 | dffb4efdef5f4e911668b6e517388332 |
|
BLAKE2b-256 | 1a5cb86669708df549c7d14ecf77b637f129ff08a99aa63e84101bdee0acfa54 |