Skip to main content

I get this error when I try to deploy Real-ESRGAN. Can anyone tell me why?

./deploy.py /home/aetina/Desktop/voyager-sdk-release-v1.4/ax_models/zoo/torch/real-esrgan-x4plus-onnx.yaml --num-cal-images 50
INFO    : Detected Metis type as m2
INFO    : Compiling network real-esrgan-x4plus-onnx /home/aetina/Desktop/voyager-sdk-release-v1.4/ax_models/zoo/torch/real-esrgan-x4plus-onnx.yaml
INFO    : Compile model: real-esrgan-x4plus-onnx
INFO    : Using representative images from /home/aetina/.cache/axelera/data/coco2017_repr400 with backend ImageReader.PIL,pipeline input color format ColorFormat.RGB
INFO    : Prequantizing real-esrgan-x4plus-onnx: real-esrgan-x4plus-onnx
INFO    : Quantizing network real-esrgan-x4plus-onnx /home/aetina/Desktop/voyager-sdk-release-v1.4/ax_models/zoo/torch/real-esrgan-x4plus-onnx.yaml real-esrgan-x4plus-onnx
INFO    : Compile model: real-esrgan-x4plus-onnx
INFO    : Using representative images from /home/aetina/.cache/axelera/data/coco2017_repr400 with backend ImageReader.PIL,pipeline input color format ColorFormat.RGB
Calibrating... ✨✨✨✨✨✨✨✨✨✨✨✨✨✨✨✨✨✨✨✨ | 100% | 39.32s/it | 50it |
Killed
Command failed: /home/aetina/Desktop/voyager-sdk-release-v1.4/deploy.py --num-cal-images 50 --model real-esrgan-x4plus-onnx --auto-vaapi --data-root /home/aetina/.cache/axelera/data --pipe gst --build-root /home/aetina/Desktop/voyager-sdk-release-v1.4/build real-esrgan-x4plus-onnx --mode QUANTIZE --metis m2  
Return code: 137
ERROR   : Failed to prequantize real-esrgan-x4plus-onnx: real-esrgan-x4plus-onnx
INFO    : Compiling real-esrgan-x4plus-onnx took 2357.972 seconds

Hi Giodst,

I am afraid you have run out of RAM, this model unfortunately needs a lot of RAM to compile - 64Gb.

If you are happy with the default training, then you can download the prebuilt model axdownloadmodel real-esrgan-x4plus-onnx


Hi Giodst, 

The Real-ESRGAN-x4plus neural network is a special case- this network requires a high amount of system RAM to compile locally, as mentioned in our release notes.

You can download the pre-compiled model for use with our SDK with the command:

axdownloadmodel real-esrgan-x4plus-onnx

 

Thanks,

David

 


Ah I understand, thanks everyone.