Machine Learning
Welcome! The Azure Machine Learning team invites you to share and vote for features to help you build, manage, and deploy custom machine learning models.
Have a technical question, or want to learn more? Please visit our documentation, MSDN forum or StackOverflow.
-
Deploy a web service to AKS without a public endpoint
It seems like the Machine Learning service currently always uses a public load balancer with a public IP when deploying a web service to AKS.
We do not want to expose our web services to the public as that is a security risk - even with the Authorization keys. Also, even if we had to expose it, we'd prefer to use our own centralized ingress system to control the traffic coming into the cluster (e.g. via Azure Application Gateway v2).
We therefore would like to be able to choose one of the following additional methods when connecting to AKS/when deploying…
41 votesThis scenario is supported now:
https://docs.microsoft.com/en-us/azure/machine-learning/how-to-enable-virtual-network#use-private-ips-with-azure-kubernetes-service.Thanks.
-
6 votes
Hi Azure Customer,
For Studio V2 (visual interface), the preinstalled pandas version is 0.2.41. And user can update to 0.2.42 by adding os.system(f"pip install pandas==0.24.2") in his script. For Studio V1, the pandas version is 0.18.0. And there is no plan to update V1.
Please refer to the document for more details: https://docs.microsoft.com/en-us/azure/machine-learning/service/ui-concept-visual-interface
Regards,
Azure CXP Community -
Support custom R modules in the Visual Interface
We have a client with a large codebase in R that we want to migrate from SPSS to Azure Machine Learning. Right now we're stuck with AML Studio which doesn't give us the performance that we need for running experiments.
9 votesHi Azure Customer,
Thank you for your feedback! Good news is, we have finished this feature request and custom R modules is supported in Visual Interface now!
Thanks again for your precious feedback!
Regards,
Azure CXP Community -
Please upgrade the version of R to R 3.5.2 used in azure
I run R codes in azure and according to my understanding it does not support the functions used in R 3.5.2 since azure uses an older version of R. Would be greatly helpful if you can upgrade the version to R 3.5.2
5 votesHi Azure Customer,
We have no plan to upgrade R version in Auzre Machine Learning Studio. We have Studio like experience in Azure Machine Learning service called “visual interface”. We just added support for “Execute R script“in visual interface. The R version is 3.5.1. We will continue to upgrade R version in visual interface. Please refer to the document: https://docs.microsoft.com/en-us/azure/machine-learning/service/ui-concept-visual-interface
Regards,
Azure CXP Community -
Please add function that can install "CLI extension for Azure Machine Learning service" at Docker Azure CLI
Please add function that can install "CLI extension for Azure Machine Learning service" at Docker Azure CLI
4 votesHi Azure Customer,
This is now fully supported through: az extension add -n azure-cli-ml Please check out aka.ms/azmlcli for more information. Thank you for your feedback!
Regards,
Azure CXP Community -
Role based access needed for Azure ML Service
We have multiple customers asking for the following role based scenarios,
- Experimentation: Define roles for users who can only submit the jobs, who can only manage computes etc.
- Inferencing: Define roles as explained below,
I’m trying to run a standard 3 env (build/test/live) set up with a single common AML Workspace (so that the immutable images tested are promoted, not re-created in a different Workspace instance).
I’d need to be able to achieve the below:
• Data Scientists can register models & create images within the AML Workspace
• Data Scientists can deploy images to ACI or Build AKS env…5 votesHi Azure Customer,
We have built this function, please refer to the document: https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-assign-roles
Thank you for your feedback ^^
Regards,
Azure CXP Community -
auto Machine learning
develop a simple solution that, I just need to upload dataset and choose few target variables. or features.
then your software will apply all the ML algorithms /with all the parameters available
and then give us best machine learning model.** ease of use, simple
1 votecompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
Hello, We have GA’ed AML service and Automated ML is a critical part of the offering. It will be great if you can review the current version of AML service and see if this meets your needs and let us know how we can improve the product better for your scenarios.
Thanks
https://azure.microsoft.com/en-us/blog/azure-machine-learning-service-a-look-under-the-hood/
-
Provide a more in-depth timeline for Azure ML Service development.
Would be nice to have a better idea when ML Service is expected to be fully released. Thanks!
2 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have GA’ed on December 4th.
https://azure.microsoft.com/en-us/blog/azure-machine-learning-service-a-look-under-the-hood/
-
Export trained MAMLS models to standalone Runtime environment
Need a way to run trained MAMLS models outside of MAMLS. Need a way to export MAMLS trained models to an external runtime environment.
3 votesHi Azure Customer,
The model trained in Azure Machine Learning Studio can’t be used outside Azure ML Studio. There is no plan to covert a MAMLS models to ONNX.
On the other hand, we have developed visual interface for Azure Machine Learning service and launch public preview in May. The new visual interface is pretty much the same user experience as Azure ML Studio. The model trained in visual interface can used in other environment.
Thank you for your understanding and feedback.
Regards,
Azure CXP Community -
Improve handling of encodings in Execute R Script
I imported some CSV files in R Script with scan. The files were encoded in UCS2-LE. In R this show as "unknown".
Otherwise all went well but in the end I got an error message saying that there was an unsupported encoding in the dataset.I solved the problem by explicitly encoding the data to UTF-8:
for (j in (1:ncol(data.set))) {
if (is.factor(data.set[, j])) {
levels(data.set[, j]) <- unlist(lapply(levels(data.set[, j]), enc2utf8))
} else {
for (i in (1:nrow(data.set))) {
if (typeof(data.set[i, j]) == "character") data.set[i, j] <- enc2utf8(data.set[i, j])
}
}
}Not pretty, but you get the idea.
I…
1 votecompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
This should work now – can you try and if it doesn’t work, please raise it in forum – https://social.msdn.microsoft.com/forums/azure/en-US/home?forum=MachineLearning
-
Provide API to read rules learned from decision trees to incorporate into application code
Use what is learned within decision trees to change application program code to drive process outcomes
1 votecompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
You should be able to create a predictive experiment from your decision tree training experiment and publish as web service. This will allow users to call API that is using learnt rules to score
-
4 votescompleted ·
Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
Latest python and R distributions have been added
-
Allow modules to run more than 24h
Currently there seems to be an undocumented limit to the runtime of a module. It would be nice to be able to configure or remove this limit per webservice or workspace
1 votecompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have increased timeout to be 48 hours
-
Get machine learning experiments under source control
It should be awesome if we could export (test) experiments and put them under source control in for example TFS, while also being able to import these experiments in a new (production) workspace.
15 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have added support for import and export web service definition and experiment JSON through PowerShell and store in version control repositary
Web Service – msdn.microsoft.com/en-us/library/azure/mt76..
Experiment – github.com/hning86/azuremlps -
Can you please add xgboost library to the existing list.
Can you please add xgboost library to the existing list.
40 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have added support for MRO 3.2.2 for execute R module which has this package
-
Add Version Control to Azure ML
Sorry if it exists already. But couldn't find a way to maintain succesive versions of models, be able to compare them with previous ones, rollback to a previous one, see a version history, etc.
10 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have added support for import and export web service definition and experiment JSON through PowerShell and store in version control repositary
Web Service – https://msdn.microsoft.com/en-us/library/azure/mt767952.aspx
Experiment – https://github.com/hning86/azuremlps -
Allow access to blob storage (and Azure, in general) from Azure ML experiments (aka: import the azure Python module)
Azure ML Experiments provide ways to read and write CSV files to Azure blob storage through the Reader and Writer modules. However, I need to write a JSON file to blob storage. Since there is no module to do so, I'm trying to do so from within an Execute Python Script module.
However, even though nearly every conceivable data science Python package is loaded (through Anacondas), no Azure libraries are loaded. Thus, we cannot access Azure blob storage (or any other Azure resource) inside an Execute Python Script module.
This is important since the existing modules to read and write…
6 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
Our new distributions in python has Azure Python SDK included in it
-
SAVE AS TRAINED MODEL
WHY SAVE AS TRAINED MODEL OPTION IS DISABLED ???
1 votecompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
This is working fine – please let us know on forum if you still are observing this issue – https://social.msdn.microsoft.com/forums/azure/en-US/home?forum=MachineLearning
-
Please upgrade R version in Azure ML Studio.
Please upgrade R version in Azure ML Studio. It is very inconvenience use the latest R packages.
23 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have added MRO 3.2.2 support.
-
copying experiments from one workspace to another should also copy descriptions
When you copy one experiment from one workspace to another workspace, it copies all modules and its comments and its summary, but the experiment description is lost, which should not.
Also copying experiments that use same datasets makes multiple copies of that dataset in the datasets of other workspace.
This would really help save extra effort of writing descriptions and deleting the sparse datasets7 votescompleted ·Adminmluservoice (Program Manager, Microsoft Azure Machine Learning, Microsoft Azure) responded
We have added support for annotations. So on experiment canvas, right click and add comment. This comment will be copied across workspaces.
- Don't see your idea?