search

ioam / topographica

A general-purpose neural simulator focusing on topographic maps.
topographica.org
BSD 3-Clause "New" or "Revised" License
53 stars 32 forks source link

gcal ipython tutorial an FixedPoint #632

Open sarahcattan opened 8 years ago

sarahcattan commented 8 years ago

I download the git version on 20 september. I tried to run the gcal model with topographica -i and topographica -n. In both, when i load the model i get: WARNING:root:Time: 000000.00 main: Casting type 'FixedPoint' to Fraction.fraction In the notebook gcal collector tutorial it came when entering: topo.sim.model = ModelGCAL(cortex_density=47) in command line it came after: runscript(model_name) I also tried to display figures after running 10000 simulations (using topographica in command line): data = c(times=[10000]) The traceback go through holoviews/interface/collector.py and holoviews/core/ndmapping.py to finish in misc/fixedpoint.py with: can't convert to FixedPoint: None I have no idea where theses None values come from and what the Fraction.fraction in the first warning means.

sarahcattan commented 8 years ago

I solved it, it was not due to topogaphica. If you are working with debain wheezy, this file /lib/x86_64-linux-gnu/libc.so.6 need to be updated to import gmpy in python (FixedPoint was used because gmpy was not found when starting topograhica). So i add the experimental depot in my source.list and i did apt-get -t experimental install libc6-dev to install the new version. see here the complete description http://www.random-dev.de/fix-version-glibc_2-14-found-error/

philippjfr commented 8 years ago

Glad to hear you managed to resolve it. We should probably either fix the support for FixedPoint as a fallbackback for gmpy or deprecate it.

mjabri commented 8 years ago

Sarah, Are you running GCAL_Tutorial? Can you actually run this statement in the tutorial ok? topo.sim.Retina.input_generator.anim(50)

sarahcattan commented 8 years ago

No, it give Out[10]: KeyError None And it appends each time i want an animation (more figures instead of the last time only)

mjabri commented 8 years ago

I see, so you get some patterns showing up. I don't get to see anything!! It seems the problem is not the patterngenerator anim() code but in the rendering of the dict returned by anim(). The dict seems ok as i can display the entries one by one.

philippjfr commented 8 years ago

I'll look at this now, will get back to you.

mjabri commented 8 years ago

Sarah, Unrelated but I am curious if you have tried the GUI with the examples/gcal.ty tutorial (ie topographica -g gcal.ty) and get to visualise projections after 1 iteration?

sarahcattan commented 8 years ago

No, one time i got it but it was VERY slow.

mjabri commented 8 years ago

The first iteration is somewhat slow for me too. I think it produces all the input patterns.

sarahcattan commented 8 years ago

If you want to run the GCAL model, i recommend you to use the file of the stable version of topographica, it run perfectly.

philippjfr commented 8 years ago

The first iteration is somewhat slow for me too. I think it produces all the input patterns.

Topographica now initializes the model only when you run the first step, so that's what's taking so long.

mjabri commented 8 years ago

Ok, Thanks. Does the stable version has the sparse/GPU code in it?

philippjfr commented 8 years ago

Does the stable version has the sparse/GPU code in it?

No, it doesn't unfortunately. Sorry I haven't managed to look at the GUI yet, hopefully I'll have some time in the next few days.

mjabri commented 8 years ago

That's great. Thanks.