Of course you were right, that I needed to make the eltypes a bit more 
flexible, so forwarddiff can do its magic.

With that fixed, it's still not working though. It runs for ~10 seconds, 
and then it just says:

julia> j = ForwardDiff.jacobian(phi, P)
Killed

I tried to search for "Killed" in the github repository but did not find 
anything. Any idea what the problem is? Am I trying to AD things that 
cannot be? (I have BLAS calls in my mapping for example)

On Saturday, November 28, 2015 at 11:45:26 PM UTC+1, Kristoffer Carlsson 
wrote:
>
> You are most likely trying to copy a "gradient number" into a vector that 
> only takes Floats or something.
>
> If you initialize your vector with zeros(10) or something, you should 
> instead initialize it like zeros(eltype(x), 10) where x is your input 
> vector.
>

Reply via email to