Does fastai's fit accept multiple dataset?


#1

If I have Image data put into CNN and would like to combine scalar data after Flatten().

def forward(self, x, p):
    cnn = self.model_cnn(x)
    fc = torch.cat((x,p))
    return self.layers(fc)

I wonder is there anyway to use fastai fit ? It seems accept only 1 dataset (or dataloader).


(Mark Worrall) #2

@Woodstock Did you find a solution? I think this is a pertinent question and from what I’ve read it’s not straightforward (but happy to be proved wrong) - it’s something that is trivial in Keras for instance.


#3

Yes, turned out that I need to write model from scratch.

Luckily, yinterian made a tutorial here