Evaluation of models based on mmflow

[luis2r]

I am using a model based on mmflow for the stereo challenge but I was not able to use the evaluation script provided for this challenge. So I adapted my evaluation script based on yours but I was not able to calculate two parameters:
flops = FlopCountAnalysis(model, inputs)
activations = ActivationCountAnalysis(model, inputs)
due to mmflow input format. Are my results acceptable without reporting flops and activations? Can you create a script to evaluate models based on mmflow?

[jhonykaesemodel]

Hi Luis, thanks for reaching out.
Yes, please submit your results without flops and activations for now. We'll consider it.
I'll take a look at mmflow in the future and see if we can adapt to it.

[luis2r]

Thank you @jhonykaesemodel