Not able to load the latest version of Audio2Face app

We have been facing some issue when loading the Audio2Face 2022.1.0 application.
It continuously gives the following error:

2022-08-01 17:59:35 [186,117ms] [Error] [omni.gpucompute-cuda.plugin] cudaMemcpyAsync failed to copy 738240 bytes from 0000000000000000 to 0000000000000000 - kind=3
2022-08-01 17:59:35 [186,263ms] [Error] [omni.gpucompute-cuda.plugin] CUDA error in mallocAsync: operation not supported: c:\b\w\b873e20d07f4fbd1\kit\source\plugins\omni.gpucompute-cuda\gpucompute-cuda.cpp: 123 (device 0)
2022-08-01 17:59:35 [186,263ms] [Error] [omni.gpucompute-cuda.plugin] CUDA error in mallocAsync: operation not supported: c:\b\w\b873e20d07f4fbd1\kit\source\plugins\omni.gpucompute-cuda\gpucompute-cuda.cpp: 123 (device 0)
2022-08-01 17:59:35 [186,263ms] [Error] [omni.gpucompute-cuda.plugin] CUDA error in memcpyAsync: invalid argument: c:\b\w\b873e20d07f4fbd1\kit\source\plugins\omni.gpucompute-cuda\gpucompute-cuda.cpp: 168 (device 0)

@esusantolim Is there any way to resolve this?

System Configuration:

  • no. of vCPUS: 18
  • Memory: 220
  • GPU: A10 on Azure
  • GPU Partition: 1/2
  • OS: Microsoft Windows Server 2022 DataCenter Gen2

@dkorobchenko for vis

There is a chance this might not run on A10. But let us confirm that.
We are also debugging some vgpu scenarios others are seeing issues as well.

Thanks.

This seems to be the same mallocAsync case that we recently addressed. It should be fixed on our next Hotfix release.