Microsoft Azure Batch AI Client Command-Line Tools
Microsoft Azure CLI Batch AI Module
This package is for the batchai module.
BatchAI commands are now deprecated and hidden.
Upgrade azure-mgmt-storage from 3.1.1 to 3.3.0
Upgrade pinned dependency azure-storage-blob from 1.1.0 to 1.3.1
Fix auto storage account creation failure in az batchai cluster create command
Logger output for auto-storage account creation now specifies “resource group”.
BREAKING CHANGE: ‘show’ commands log error message and fail with exit code of 3 upon a missing resource.
Fixed az batchai job exec command
Fixed -o table option for az batchai cluster node list and az batchai job node list commands.
Added support for 2018-05-01 API
Added support for workspaces. Workspaces allow to group clusters, file-servers and experiments in groups removing limitation on number of resources can be created;
Added support for experiments. Experiments allow to group jobs in collections removing limitation on number of created jobs;
It’s possible to configure /dev/shm for jobs running in docker container.
Added ‘az batchai cluster node exec’ and ‘az batchai job node exec’ commands. These commands allow to execute any commands directly on nodes and provide functionality for port forwarding. Port forwarding can be used, for example, to access tensorboard and jupyter running on cluster’s nodes;
az batchai now supports –ids parameters like other az commands;
Breaking change: now all clusters and fileservers must be created under workspaces;
Breaking change: now jobs must be created under experiments;
Breaking change: ‘–nfs-resource-group’ option is deleted from ‘cluster create’ and ‘job create’ commands. To mount NFS belonging to a different workspace/resource group provide file server’s ARM ID via ‘–nfs’ option;
Breaking change: ‘–cluster-resource-group’ option is deleted from ‘job create’ command. To submit a job on a cluster belonging to a different workspace/resource group provide cluster’s ARM ID via ‘–cluster’ option;
Breaking change: jobs, cluster and file servers do not longer have location attribute. Location now is an attribute of a workspace. So, ‘–location’ parameter has been removed from ‘job create’, ‘cluster create’ and ‘file-server create’ commands;
Breaking change: names of short options were changed to make interface more consistent:
In create cluster, file-server and job commands [’–config’, ‘-c’] option was renamed to [’–config-file’, ‘-f’];
In create job command [’–cluster’, ‘-r’] option was renamed to [’–cluster’, ‘-c’];
In ‘job file list’ and ‘job file stream’ commands [’–job’, ‘-n’] option was renamed to [’–job’, ‘-j’];
In ‘cluster file list’ command [’–cluster’, ‘-n’] option was renamed to [’–cluster’, ‘-c’]
Now ‘az batchai create cluster’ respects vm priority configured in the cluster’s configuration file.
Added support for 2018-03-01 API
Job level mounting
Environment variables with secret values
Performance counters settings
Reporting of job specific path segment
Support for subfolders in list files api
Usage and limits reporting
Allow to specify caching type for NFS servers
Support for custom images
Added pyTorch toolkit support
Added ‘job wait’ command which allows to wait for the job completion and reports job exit code
Added ‘usage show’ command to list current Batch AI resources usage and limits for different regions
National clouds are supported
Added job command line arguments to mount filesystems on the job level in addition to config files
Added more options to customize clusters - vm priority, subnet, initial nodes count for auto-scale clusters, specifying custom image
Added command line option to specify caching type for Batch AI managed NFS
Simplified specifying mount filesystem in config files. Now you can omit credentials for Azure File Share and Azure Blob Containers - CLI will populate missing credentials using storage account key provided via command line parameters or specified via environment variable or will query the key from Azure Storage (if the storage account belongs to the current subscription).
Job file stream command now auto-completes when the job is completed (succeeded, failed, terminated or deleted).
Improved ‘-o table’ support for show operations.
Added –use-auto-storage option for cluster creation. This option make it simpler to manage storage accounts and and mount Azure File Share and Azure Blob Containers to clusters.
Added –generate-ssh-keys option into ‘cluster create’ and ‘file-server create’.
Added ability to provide node setup task via command line.
Breaking change: ‘job stream-file’ and ‘job list-files’ commands are grouped under ‘job file’ group.
Breaking change: renamed –admin-user-name to –user-name in ‘file-server create’ command to be consistent with ‘cluster create’ command.
sdist is now compatible with wheel 0.31.0
Update for CLI core changes.
Added short option for providing VM size in file-server create command
Added storage account name and key arguments into cluster create parameters
Fixed documentation for job list-files and stream-file
Added short option for providing cluster name in job create command
Initial release of Batch AI module.
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Hashes for azure_cli_batchai-0.4.10-py2.py3-none-any.whl