exceptions when creating the dataset for the "tile_smoke" problem

[joseph pareti]

Here's the log file from my Ubuntu VM (in Azure). I have had several troubles with this case, presumably due to incompatible python releases. But now it appears to be consistent:

strings -a /backupdata/home/joepareti54/FluidNet/manta/build/manta | grep -i pyth

libpython3.5m.so.1.0

_ZN2Pb15WrapperRegistry13addPythonCodeERKSsS2_

_ZN2Pb15WrapperRegistry13addPythonPathERKSs

cannot allocate new python object

can't allocate new python class object

PYTHON operators have to be defined within classes.

unable to set python path

PYTHON operator

python class

# Defines some constants for use in python subprograms

python/defines.py

_ZN2Pb15WrapperRegistry13addPythonCodeERKSsS2_

_ZN2Pb15WrapperRegistry13addPythonPathERKSs

and yet the application "manta" fails, as such it cannot generate the training data. 

see http://ge.in.tum.de/publications/2017-sig-chu/

Here's the log: Any ideas?

$ which manta

/backupdata/home/joepareti54/FluidNet/manta/build//manta

$pwd

/backupdata/home/joepareti54/Muenchen/manta/tensorflow/example1_smoke_tiled

$ manta manta_genSimData.py

Version: mantaflow 64bit fp1 commit dd3bb0c0a65cc531d3c33487bde5edcb4aa6784f from Apr  5 2018, 09:17:54

Loading script 'manta_genSimData.py'

noise tile loaded from file!

0 centre [+120.806343,+92.165733,+0.500000] radius 13.644765537175845 other [+0.979857,+1.038559,+1.000000]

1 centre [+130.754028,+122.131645,+0.500000] radius 16.004152612279245 other [+0.980772,+1.003861,+1.000000]

2 centre [+98.904694,+140.872925,+0.500000] radius 14.774846027071572 other [+0.960563,+0.950266,+1.000000]

3 centre [+139.908112,+153.478745,+0.500000] radius 16.04314154211211 other [+0.953714,+1.002965,+1.000000]

4 centre [+162.714828,+98.070480,+0.500000] radius 17.79526578119472 other [+1.041082,+1.029490,+1.000000]

5 centre [+172.805069,+136.113434,+0.500000] radius 16.154283987865018 other [+1.006432,+1.001612,+1.000000]

6 centre [+112.838028,+133.982483,+0.500000] radius 12.578920662086773 other [+0.962552,+0.966262,+1.000000]

7 centre [+153.994293,+143.740540,+0.500000] radius 16.4697652927247 other [+1.002658,+1.010280,+1.000000]

8 centre [+172.400192,+126.488327,+0.500000] radius 10.243350219257 other [+0.991511,+0.990707,+1.000000]

9 centre [+108.644135,+137.781693,+0.500000] radius 15.215080431219896 other [+0.951960,+1.005932,+1.000000]

10 centre [+130.359299,+176.941513,+0.500000] radius 17.177258330136695 other [+0.960588,+0.968750,+1.000000]

11 centre [+126.819069,+119.025108,+0.500000] radius 11.732635085142041 other [+0.966099,+0.989244,+1.000000]

Build info: mantaflow 64bit fp1 commit dd3bb0c0a65cc531d3c33487bde5edcb4aa6784f from Apr  5 2018, 09:17:54

Using sim dir '../data/sim_1001/'

Current time t: 0.0

Traceback (most recent call last):

  File "manta_genSimData.py", line 188, in <module>

    newCentre = calcCenterOfMass(xl_density)

NameError: name 'calcCenterOfMass' is not defined

Script finished.

$ strings -a /backupdata/home/joepareti54/FluidNet/manta/build/manta | grep -i ofmass

_ZNK5Manta4Mesh19computeCenterOfMassERNS_8Vector3DIfEE

_ZNK5Manta4Mesh19computeCenterOfMassERNS_8Vector3DIfEE

--

Regards, 

Joseph Pareti - Artificial Intelligence consultant

cell +49 1520 1600 209

cell +39 339 797 0644

[Nils Thuerey]

Hello Joseph, which mantaflow version are you compiling? calcCenterOfMass is defined in initplugins.cpp:

source/plugin/initplugins.cpp: PYTHON() Vec3 calcCenterOfMass(const Grid<Real>& density)

It looks like you're using the FluidNet mantaflow version, which won't work. Better run the data generation with the corresponding version, e.g., mantaPatch for the Chu 2017 paper.

Cheers, Nils

[joseph pareti]

Indeed, thanks Nils. I was stuck for a long time working with the fluidnet version. After rebuilding a training I got a reasonable ouptut. See attached a couple of diagrams, calculated and expected.

regards

--
You received this message because you are subscribed to the Google Groups "mantaflow Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mantaflow-forum+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mantaflow-forum/2f9e0eeb-4c6c-4f7f-b275-4ca953f665cc%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Nils Thuerey]

Nice, that's good to see!

[joseph pareti]

thanks. Juts a clarification on the ouptut files:

calculated: what the deep network predicts

expected: what comes out of the accurate CFD calculation

Am I right?

--
You received this message because you are subscribed to the Google Groups "mantaflow Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mantaflow-forum+unsubscribe@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/mantaflow-forum/5f99bc2b-77f5-4cae-b92c-21afb1158a95%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[joseph pareti]

here's a status report on my work in Azure. As explained in the test I ran the whole chain twice,  and got a different number of output images in the 2 runs, which I don't understand.

Regards,

[Nils Thuerey]

Regarding - your previous question, it's a bit hard to tell without context, but yes, "expected" typically is the ground truth solution.

For the different numbers of outputs, it probably depends on how you called the scripts each time?

Cheers, Nils