elQ model conversion - NOT image classification

[colinbroderick]

Can the elQ toolkit be used to convert models which are NOT used for image classification? We have a model that produces a continuous output and there is therefore no sensible concepts of labels, but the conversion will not succeed without providing a labels file, at least in the graphical tool.

Can anyone offer advice or resources for workflows that do NOT involve image classification? It is frustratingly hard to make progress because virtually all training resources and documents and example code assume the image classification case.

In particular, at the moment, we're most interested in converting from PyTorch to ONNX. We can do this with other tools but I was keen to try out the quantization capabilities of the elQ toolkit.

Thanks

[Bio_TICFSL · Accepted solution]

Hello colinbroderick,

The eIQ support Pytorch and ONNX, but it can not be used for model conversion it they are not going to use in Image classification, for this task you have the Linux BSP, where you can install an apps to do this task.

Regards

[colinbroderick]

Ok, thanks for the answer.