Just thought I mention that I've reproduced this bug again on my laptop. Below some output, it keeps repeating at infinitum (got as far as level l=14021)...
This is for just the minimize_nested_blockmodel_dl(), no layers, no weights, bipartite structure. 
Hope this helps,
Peter

    l=1  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=1  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=1  Bisect at B = 2 with S = 31.31709660208764

    l=1  Best result: B = 2, S = 31.31709660208764

level 1 : rejected replacement (2, 1) -> (2, 2) , dS: nan

    l=1  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=1  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=1  Bisect at B = 2 with S = 31.31709660208764

    l=1  Best result: B = 2, S = 31.31709660208764

level 2 : inserted 2 , dS: nan

    l=3  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=3  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=3  Bisect at B = 2 with S = 31.31709660208764

    l=3  Best result: B = 2, S = 31.31709660208764

level 3 : rejected replacement (2, 1) -> (2, 2) , dS: nan

    l=3  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=3  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=3  Bisect at B = 2 with S = 31.31709660208764

    l=3  Best result: B = 2, S = 31.31709660208764

level 4 : inserted 2 , dS: nan

    l=5  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=5  Current bracket: (2, 2, 2) (31.317096602087641, 31.317096602087641, 31.317096602087641)

    l=5  Bisect at B = 2 with S = 31.31709660208764

    l=5  Best result: B = 2, S = 31.31709660208764


On Thu, 19 Jan 2017 at 08:52 Peter Straka <p.straka@unsw.edu.au> wrote:
Great, thanks Tiago

On Wed, 18 Jan 2017 at 20:06 Tiago de Paula Peixoto <tiago@skewed.de> wrote:
On 18.01.2017 01:36, Peter Straka wrote:
> Dear list,
>
> I'm trying to fit a nested blockmodel to a (bipartite) network with ~10^6
> edges. The algorithm minimize_nested_blockmodel_dl() doesn't terminate; it
> keeps on adding layers indefinitely until it runs out of memory / hits
> walltime, see below. What could be going on?

This is a silly bug due to finite machine precision. It has already been
fixed in git, and will be available in the next release.

Best,
Tiago

--
Tiago de Paula Peixoto <tiago@skewed.de>
_______________________________________________
graph-tool mailing list
graph-tool@skewed.de
https://lists.skewed.de/mailman/listinfo/graph-tool
--
Dr Peter Straka
Research Fellow (DECRA)
School of PEMS | UNSW Canberra
T: +61 (2) 9385 7024 | +1 313 757 0137