Closed elad-bar closed 1 year ago
According to Vision/modulesettings.docker.json, face recognition processing does not support CUDA:
{
"Modules": {
"FaceProcessing": {
"Runtime": "python38",
"EnvironmentVariables": {
"USE_CUDA": "false"
}
},
"SceneClassification": {
"Runtime": "python38"
},
"VisionObjectDetection": {
"Runtime": "python38"
}
}
}
Tried to load the image of GPU, modules are being loaded throughout the GPU (
nvidia-smi
command), while processing is being done on the CPU.