Plot top losses problem?

Part 1 (2020)
12

I still have the same issue.
I fixed it with this:

I redefined the function:

def plot_top_losses_fix(interp, k, largest=True, **kwargs):
        losses,idx = interp.top_losses(k, largest)
        if not isinstance(interp.inputs, tuple): interp.inputs = (interp.inputs,)
        if isinstance(interp.inputs[0], Tensor): inps = tuple(o[idx] for o in interp.inputs)
        else: inps = interp.dl.create_batch(interp.dl.before_batch([tuple(o[i] for o in interp.inputs) for i in idx]))
        b = inps + tuple(o[idx] for o in (interp.targs if is_listy(interp.targs) else (interp.targs,)))
        x,y,its = interp.dl._pre_show_batch(b, max_n=k)
        b_out = inps + tuple(o[idx] for o in (interp.decoded if is_listy(interp.decoded) else (interp.decoded,)))
        x1,y1,outs = interp.dl._pre_show_batch(b_out, max_n=k)
        if its is not None:
            #plot_top_losses(x, y, its, outs.itemgot(slice(len(inps), None)), L(self.preds).itemgot(idx), losses,  **kwargs)
            plot_top_losses(x, y, its, outs.itemgot(slice(len(inps), None)), interp.preds[idx], losses,  **kwargs)
        #TODO: figure out if this is needed
        #its None means that a batch knows how to show itself as a whole, so we pass x, x1
        #else: show_results(x, x1, its, ctxs=ctxs, max_n=max_n, **kwargs)

And using the command below I can now plot the top losses:

plot_top_losses_fix(interp, 10, nrows=2)

Hope this helps for the people who have still the same problem.

1 Like