| Re: [sumo-user] question about the --seed parameter |
|
Hello Jakob,
Thank you against for the answer. It means that, it not possible to get some deterministic value by using Python tool “createVehTypeDistribution.py”, if just one value inside the Entry files
change. Even if the seed parameter is a constant value. The only one possibility is to have the same Exit File is to you the exact same Entry File and a constant seed value.
I try to make Sumo crash against by changing the seed value, and then I would give you the Input File for this case. I saw that the 2 issue bellow is already closed. It means that I can already use the last release of yesterday? Or should I wait tomorrow? - the issue with collisions is due to safety issues with the new car following model ACC (https://github.com/eclipse/sumo/issues/4551) - there are further collisions triggered by problems with the opposite-direction model. Some of them have been fixed in the meantime but some still remain. (https://github.com/eclipse/sumo/issues/3718)
Br Christian
Von: sumo-user-bounces@xxxxxxxxxxx [mailto:sumo-user-bounces@xxxxxxxxxxx]
Im Auftrag von Jakob Erdmann The python tool createVehTypeDistribution.py uses a single random number generated. For this reason, changes to some parameters also affect the values for other parameters.
Determinism is only guaranteed insofar the exact same inputs will result in the exact same results when using the same seed. If you could provide the exact input files that make sumo crash, I'd be happy to investigate (crashes are not supposed to happen). regards, Jakob 2018-09-12 17:13 GMT+02:00 <christian.damdjowabo@xxxxxx>: Hello Jakob, hello everybody,
I have a question about the generation distribution using “createVehTypeDistribution.py” or also by generation the road using “duarouter”. My question is that I don’t know why by changing just one parameter (as example the parameter “mean impatience”) and keep another parameter constant.
The generated value isn’t constant, although I use a constant value of “--seed” (--seed=1) as you can see in the command line that I used.
To be able to explain you exactly the problem that I facing now, I sent you some files which contain the “typ1.txt” and “typ2.txt”.
I generated the first time “vTypeDistributions1.add.xml” and “vTypeDistributions2.add.xml”. In this first generation the mean value of parameter “impatience” is equal to 0.
And then I changed in the second generation that value to “1” and the other parameters are always kept constant.
Unfortunately I got in the “vTypeDistributions1.add.xml” and “vTypeDistributions2.add.xml” a change in all parameter. Also the parameter, which was kept constant in the Input File
“typ1.txt” and “typ2.txt”. Can you help me to know, what I’m doing wrong. Because I through that by setting the parameter --seed = constant, it would be possible to have a deterministic choice of random parameter.
I don’t know why it’s not the case by using “createVehTypeDistribution.py”. I also remark that for some value of --seed sumo crashes. It’s why I want to know, how we have to choice that value?
--------------------- Part of the code below--------------------------------------------------------------------------------------------------|
commandStr1 =
'python D:\criticality_pyth\createVehTypeDistribution.py D:\criticality_pyth\typ1.txt --output-file=D:\criticality_pyth\vTypeDistributions1.add.xml --name=vehDist1 --size=1000
--seed=1';
commandStr2 =
'python D:\criticality_pyth\createVehTypeDistribution.py D:\criticality_pyth\typ2.txt --output-file=D:\criticality_pyth\vTypeDistributions2.add.xml --name=vehDist2 --size=1000
--seed=1';
commandStr3 =
'python D:\criticality_pyth\randomTrips.py --seed=1 --weights-prefix=D:\criticality_pyth\example.src.xml --prefix=pa --period=3 --net-file=D:\criticality_pyth\Campus_test.net.xml
--route-file=D:\criticality_pyth\campus_test_passenger.rou.xml --trip-attributes="type=\"vehDist1\"" --additional-file=D:\criticality_pyth\vTypeDistributions1.add.xml';
commandStr4 =
'duarouter D:\criticality_pyth\flows2routes.duarcfg --seed=1'; Br
Christian
|