Open pgleeson opened 7 months ago
@pgleeson I keep getting this error and I don't understand why? "RuntimeWarning: invalid value encountered in scalar divide ret = ret.dtype.type(ret / rcount)"
I think it might have something to do with the range of values I'm inputting, but I'm not sure. Do you know what it could be?
Difficult to tell without looking at the code. Which file is it and which model are you trying it on?
It's the generate_current_vs_frequency_curve function that seems to be doing this. It's doing it for every model (including the Hay and Bahl models) and I can't understand why. It was working fine on Saturday and I don't know what I did that caused this problem.
Can you give us the exact command/code snippet you're running please? We can't debug this if we can't reproduce it. So generally, you should tell us:
pip list
With this info, we can do exactly what you did and see if we get the warning too, and if we do, we'll be in a position to dive into the code to see what's going on.
If you commit small changes to your scripts etc. frequently to git, you will also be able to pin point exactly when the issue began and then see what you did that began it.
My bad Ankur, I should have mentioned this. I sorted out this problem on Tuesday when I came into the office. It was a combination of random error messages that didn't affect the end result, a bug in the code that makes the markdown file and a lack of available RAM during the simulation.
I made the mistake of not fully testing the code, before pushing to GitHub, which caused this whole mess. Thank you for the tips, I will make sure to add a screenshot of the errors next time there is a problem and make commits more carefully and more regularly.
No worries, as long as it got sorted :+1:
I just pushed the combined IF graph to the dev branch.
@pgleeson I'm a bit confused. The above combined IF graph considers each cell as an individual. Would this be sufficient or do I need something from the Yao model working as one?
Re your image above, looks good, but it would be nice if the colours matched those used in Yao.
For now, we can assume that all this will tell us is the average firing rate of all cells in a disconnected population of neurons of each type, i.e. a population of N SST cells would fire at an average of 70Hz if the population was uniformly injected with a current of 300pA, independent of N. That can be the basis for setting the individual rate based population parameters in del Molino for the case where all weights are zero...
These are the latest NeuroML versions of the human cell models: https://github.com/sanjayankur31/Human-L2-3-Cortical-Microcircuit/blob/feat/neuroml/NeuroML2/HL23PYR.cell.nml https://github.com/sanjayankur31/Human-L2-3-Cortical-Microcircuit/blob/feat/neuroml/NeuroML2/HL23PV.cell.nml https://github.com/sanjayankur31/Human-L2-3-Cortical-Microcircuit/blob/feat/neuroml/NeuroML2/HL23SST.cell.nml https://github.com/sanjayankur31/Human-L2-3-Cortical-Microcircuit/blob/feat/neuroml/NeuroML2/HL23VIP.cell.nml
Note that it's not on the main branch of that repo, but https://github.com/sanjayankur31/Human-L2-3-Cortical-Microcircuit/tree/feat/neuroml_pg
The VIP cell is the smallest, so should be best for testing as it will run faster