Got wrong answer when converting from torch7

1

I have converted a model from crnn in torch7 (https://github.com/bgshih/crnn), but the predictions are completely wrong. How can I debug it?
I want to print the tensor value in each layer in torch7 and compare it with pytorch, but it seems that there is no print layer in torch7. Now I am trying to load the model in torch and remove some layers to get the output. Is there any better idea?

2

Did you make sure to set your model to evaluation with model.eval() before comparing its output to the torch7 model?
I think in torch7 you can just print the output of each layer.
Could you try to call this, if you used a Sequential model:

net.modules[n].output
3

It turns out my input is not correct, thanks!