Is it possible to call any ONNX Runtime Execution provider through WinML API? I'm able to run custom DML through WinML but unable to find steps for WinML to onnxruntime EP.
thanks for the question. The supported path for the Windows ML API is to use the DML EP. This diagram illustrates how Windows ML uses the ONNX Runtime for CPU inferencing, and the DirectML EP for GPU acceleration across all Windows devices.
Related
I am working on a predictive on IBM SPSS modeler with an SVM and then uploading the PMML to the cloud to provide it as a service, but even when I use the automated system to send the request it gives me ArrayOutOfBounds exceptions. Please let me know if you need more details
I am unable to deploy analysis service tabular model from VS 2019. The error message returned is "1500 is not a valid value for this element.". Also I checked the deployment analysis server in SSMS to find the supported compatibility levels does not include 1500 (the maximum level it supports is 1480 and the default one is 1200). I'm not sure if it has to do with the issue. Could someone be able to help? Thanks.
It looks like you are trying to deploy a tabular model with a compatibility mode of 1500 (The latest version of Tabular) to an instance of that is will only support a compatibility mode of 1480, which I think is a SQL Server 2019 Analysis Services CTP2.X instance. Here are some details of the compatibility modes options and how to set them
You will have to do one of two things. But it will be dependent on the features used in the Data Model.
If you are using for example an SSAS 2019 feature like calculation groups in your data model you will have to update the SSAS instance from the CTP to the RTM release.
If you are not using any features in your data model that are compatibility mode 1500, you can set your compatibility mode to 1400 in your data model
Hope that helps
I was using MATLAB simulink to run some examples on FRDM-KL25Z board and I was unable to deploy any of the examples into the hardware.
One of the examples prompted me error and I have attached along with the attachment.
Do I need to install any third-party software such as Kinetis Design Studio to deploy the library to the board? Error File
We've run into an issue with the New-Deployment Azure Powershell commandlet timing out; we've put in a bug report with MS. While they gave us an explanation for it (the path and timeout threshold used to upload through commandlets is different then what's used by the web portal); they don't have a fix for us.
We need to get this running so we can automate our build deployments, so we're looking into developing a custom commandlet to replace New-Deployment using the Azure SDK; hoping this path will not have the timeout issues the commandlet did. But before we go down that route; are there any other scriptable tools I can use to replace the New-Deployment functionality? I looked at Cloudberry for Windows Azure; but that doesn't have a scriptable interface yet.
Any constructive input is greatly appreciated.
If you are developing worker roles, then you can reuse the dynamic assembly loading system that comes with the execution framework of Lokad.Cloud.
Basically, you just have to upload a ZIP archive containing all your DLLs toward the Blob Storage and the framework takes care of the rest. Extra Bonus: worker restart takes about 10s (vs +10min for a classical deployment)
A similar behavior could be obtained with web roles too, but this part hasn't been implemented in Lokad.Cloud yet.
FYI, we solved this issue by writing a custom tool to upload the package with a higher timeout tolerance; as well as using CSManage (http://code.msdn.microsoft.com/Release/ProjectReleases.aspx?ProjectName=windowsazuresamples&ReleaseId=3233) to create the deployment after uploading.
I am currently working on a project in Microsoft Robotics Development Studio, and I am having a little trouble deploying the solution for general use.
My problem is this. I am making a simulation, so I need to deploy the simulation engine along with my project. This is dependent on the Ageia PhysX engine, DirectX, and the Microsoft XNA framework. I understand this, and I understand that the deployment tool for MRDS (dssdeploy) isn't set up to deploy simulations by default. So, if anyone has worked with MRDS before, and has deployed a simulation, I could use some help with this.
Unfortunately, dssdeploy does not work with simulation projects. It won't package up all the required simulation libraries. I believe the reason for this is due to license and/or install issues for PhysX, DirectX, and XNA. I think you can still zip up your own service with dssdeploy, but you will have to have your target user install RDS first, then un-deploy your service.
FYI, because of this and many other headaches using dssdeploy, I have stopped using it to distribute my code. I simply zip up my service, then tell the end user to run dssprojectmigration on the code after they un-zip it.