When having a computation graph with multiple outputs of different sizes how would I call fit(INDArray[], INDArray[])?
No I mean I have multiple output layers. One with outputshape [1880] and one with [1].
That is not a problem. The second argument to the fit method is an array too. So it can take multiple INDArrays.
Then I get an error Labels and preOutput must have equal shapes: got shapes [1] vs [1, 1]
My input array is the one INDArray with the input and my labes array is the array with the two INDArray of shape [1880] and [1]
What is the shape of the labels? They should be 1x1880 and 1x1. Alternatively what happens when you give it a batchsize > 1?
My labels are of shape
Rank: 1, DataType: FLOAT, Offset: 0, Order: c, Shape: [1880], Stride: [1]
Rank: 1, DataType: FLOAT, Offset: 0, Order: c, Shape: [1], Stride: [1]
Yeah reshape them to be rank == 2 with leading dimension = 1. You have to specify the minibatch even if it’s size 1…If that turns out to be the problem, you can file an issue as enhancement.
Ok thank you! 