Closed HilbertHuangHitomi closed 3 years ago
espenhgn
commented
3 years ago Hi. I think I need more info in order to try and reproduce this crash. Can you post the output of conda info and conda list?
Which what is the git SHA for your NEST installation? When starting NEST (nest on the command line may print out Version: master@8f5a5fc1e or similar).
espenhgn
commented
3 years ago Hi again. If you're using conda my recommendation is to use as many dependencies as possible from conda-forge in order to avoid issues with slightly incompatible versions of gcc etc. between NEST and NEURON. Whatever was used to compile NEURON for PyPi.org may be different to what you used with NEST compiled locally.
I could run the simplest case example_brunel.py by the following procedure:
conda create -n hybridlfpy python=3.8 lfpy libtool gsl doxygen cmake -c conda-forge
conda activate hybridlfpy
# compile nest
git clone https://github.com/nest/nest-simulator.git
cd nest-simulator
git checkout 8f5a5fc1e # this is the version I've been using. Things may have changed upstream lateron.
mkdir nestbld
cd nestbld
cmake -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX -Dwith-mpi=ON ..
make -j4
make install
cd ../..
cd hybridLFPy
<git checkout nest3> # make sure to use the origin nest3 branch
python setup.py install
cd examples
pip install git+https://github.com/NeuralEnsemble/parameters
nrnivmodl
mpirun -np 2 python example_brunel.pyshould then hopefully work.
Also, please keep in mind that the microcircuit examples are perhaps best executed on HPC resources due to long simulation times on typical laptops/desktops. On HPC/compute clusters using conda may not be the best idea as you're better off most likely to compile both NEURON and NEST using the system compilers and libraries which may be better optimized.
HilbertHuangHitomi
commented
3 years ago All works well following your suggestions. Thank you very much! (≧∇≦)ﾉ
espenhgn
commented
3 years ago Great!
Hi~ I would like to use
cellsim16pops_*.pyto simulate. It works well during the LIF network simulation and processing the raw nest output, but fails to simulate the multi-compartment models.Here is my scheme.
conda create --name neuro python=3.8LFPy==2.1.2,neuron==7.8.2,hybridLFPy==0.1.4nrnivmodlmpiexec python cellsim16pops_default.pyormpirun -np 256 python cellsim16pops_default.pyBoth the LIF network simulation and the post-processed part of the multi-compartment simulation works well and gives expected results to the folder.
However, after reporting
population initialized in X seconds, it raises Segmentation fault asI have tried to fix this with install neuron 8.0a following https://github.com/neuronsimulator/nrn/pull/1010, downloading the wheel and running
pip install NEURON-8.0a0-cp38-cp38-manylinux1_x86_64.whlin a new environment, but the error still rises.Could you help me to solve it, please?