Helveg
commented
8 months ago Can you post the code that led to this error, so I can reproduce it?
Open danilobenozzo opened 8 months ago
Helveg
commented
8 months ago Can you post the code that led to this error, so I can reproduce it?
danilobenozzo
commented
8 months ago import logging
import bluepyopt as bpop
import bluepyopt.ephys as ephys
#import dbbs_models
from bsb.morphologies import parse_morphology_file
from arborize import define_constraints
#from arborize.optimizers._bluepyopt import get_bpo_cell
from arborize.builders._bluepyopt import bluepyopt_build
from arborize.schematics import bsb_schematic
logging.getLogger().setLevel(logging.DEBUG)
def create_schema(constraints, tags, file_morphology):
_m = parse_morphology_file(file_morphology, tags=tags)
return bsb_schematic(_m, constraints)
schema = create_schema(
define_constraints(
{
"synapse_types": {
("AMPA", "granule"): {
"tau_facil": 5,
"tau_rec": 8,
"tau_1": 1,
"gmax": 140000,
"U": 0.43,
},
("NMDA", "granule"): {
"tau_facil": 5,
"tau_rec": 8,
"tau_1": 1,
"gmax": 23500,
"U": 0.43,
},
("GABA", "granule"): {"U": 0.35},
},
"cable_types": {
"soma": {
"cable": {"Ra": 100, "cm": 2},
"ions": {"k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}},
"mechanisms": {
"Leak": {
"e": -60,
"gmax": [0.00029038073716, 0.00029038073717],
},
"Kv3_4": {"gkbar": 0.00076192450952},
"Kv4_3": {"gkbar": 0.00281496839061},
"Kir2_3": {"gkbar": 0.00074725514702},
"Ca": {"gcabar": 0.00060938071784},
"Kv1_1": {"gbar": 0.00569738264555},
"Kv1_5": {"gKur": 0.00083407556714},
"Kv2_2": {"gKv2_2bar": 1.203410852e-05},
("cdp5", "CR"): {},
},
},
"dendrites": {
"cable": {"Ra": 100, "cm": 2.5},
"ions": {"k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}},
"mechanisms": {
"Leak": {"e": -60, "gmax": 0.00025029700737},
("Leak", "GABA"): {},
"Ca": {"gcabar": 0.00500128008459},
"Kca1_1": {"gbar": 0.01001807454651},
"Kv1_1": {"gbar": 0.00381819207934},
("cdp5", "CR"): {},
},
},
"axon": {"cable": {}, "ions": {}, "mechanisms": {}},
"ascending_axon": {
"cable": {"Ra": 100, "cm": 1},
"ions": {
"na": {"rev_pot": 87.39},
"k": {"rev_pot": -80.993},
"ca": {"rev_pot": 137.5},
},
"mechanisms": {
("Na", "granule_cell"): {"gnabar": 0.02630163681502},
"Kv3_4": {"gkbar": 0.00237386061632},
"Leak": {"e": -60, "gmax": 9.364092125e-05},
"Ca": {"gcabar": 0.00068197420273},
("cdp5", "CR"): {},
},
},
"parallel_fiber": {
"cable": {"Ra": 100, "cm": 1},
"ions": {
"na": {"rev_pot": 87.39},
"k": {"rev_pot": -80.993},
"ca": {"rev_pot": 137.5},
},
"mechanisms": {
("Na", "granule_cell"): {"gnabar": 0.01771848449261},
"Kv3_4": {"gkbar": 0.00817568047037},
"Leak": {"e": -60, "gmax": 3.5301616e-07},
"Ca": {"gcabar": 0.0002085683353},
("cdp5", "CR"): {},
},
},
"axon_initial_segment": {
"cable": {"Ra": 100, "cm": 1},
"ions": {
"na": {"rev_pot": 87.39},
"k": {"rev_pot": -80.993},
"ca": {"rev_pot": 137.5},
},
"mechanisms": {
("Na", "granule_cell_FHF"): {"gnabar": 1.28725006737226},
"Kv3_4": {"gkbar": 0.00649595340654},
"Leak": {"e": -60, "gmax": 0.00029276697557},
"Ca": {"gcabar": 0.00031198539472},
"Km": {"gkbar": 0.00056671971737},
("cdp5", "CR"): {},
},
},
"axon_hillock": {
"cable": {"Ra": 100, "cm": 2},
"ions": {
"na": {"rev_pot": 87.39},
"k": {"rev_pot": -80.993},
"ca": {"rev_pot": 137.5},
},
"mechanisms": {
"Leak": {"e": -60, "gmax": 0.0003695818972},
("Na", "granule_cell_FHF"): {"gnabar": 0.00928805851462},
"Kv3_4": {"gkbar": 0.02037346310915},
"Ca": {"gcabar": 0.00057726155447},
("cdp5", "CR"): {},
},
},
},
},
use_defaults=True,
),
{
16: ["axon", "axon_hillock"],
17: ["axon", "axon_initial_segment"],
18: ["axon", "ascending_axon"],
19: ["axon", "parallel_fiber"],
},
r"/home/benozzo/dbbs_lab/repos/models/dbbs_models/morphologies/GranuleCell.swc",
)
#cell = get_bpo_cell(schema)
cell = bluepyopt_build(schema)
soma_loc = ephys.locations.NrnSeclistCompLocation(
name="soma", seclist_name="soma", sec_index=0, comp_x=0.5
)
NUMBER_INDIVIDUALS = 1 # Number of individuals in offspring
NUMBER_GENERATIONS = 1 # Maximum number of generations
sweep_protocols = []
for protocol_name, amplitude in [
# ("step1", 0.01),
("step2", 0.016),
# ("step3", 0.022),
]:
stim = ephys.stimuli.NrnSquarePulse(
step_amplitude=amplitude,
step_delay=100,
step_duration=100,
location=soma_loc,
total_duration=200,
)
rec = ephys.recordings.CompRecording(
name="%s.soma.v" % protocol_name, location=soma_loc, variable="v"
)
protocol = ephys.protocols.SweepProtocol(protocol_name, [stim], [rec])
sweep_protocols.append(protocol)
threestep_protocol = ephys.protocols.SequenceProtocol(
"twostep", protocols=sweep_protocols
)
# NEURON sim
nrn = ephys.simulators.NrnSimulator(dt=0.025, cvode_active=False)
# feature of obj function
efel_feature_means = {
"step1": {
"AP_height": 20.93,
"ISI_CV": 0.261,
"AHP_depth_abs_slow": -52.69,
"AP_width": 0.665,
"voltage_base": -68.5,
"AHP_depth_abs": -59.21,
"time_to_first_spike": 31.9,
"adaptation_index2": 0.1062,
"mean_frequency": 25,
},
"step2": {
"AP_height": 19.255,
"ISI_CV": 0.14,
"AHP_depth_abs_slow": -48.935,
"AP_width": 0.695,
"voltage_base": -68.77,
"AHP_depth_abs": -58.3,
"time_to_first_spike": 19.0,
"adaptation_index2": 0.034,
"mean_frequency": 40,
},
"step3": {
"AP_height": 17.645,
"ISI_CV": 0.148,
"AHP_depth_abs_slow": -32.67,
"AP_width": 0.7135,
"voltage_base": -69.125,
"AHP_depth_abs": -57.191,
"time_to_first_spike": 14.65,
"adaptation_index2": 0.029,
"mean_frequency": 50,
},
}
# obj function
objectives = []
for protocol in sweep_protocols:
stim_start = protocol.stimuli[0].step_delay
stim_end = stim_start + protocol.stimuli[0].step_duration
for efel_feature_name, mean in efel_feature_means[protocol.name].items():
feature_name = "%s.%s" % (protocol.name, efel_feature_name)
feature = ephys.efeatures.eFELFeature(
feature_name,
efel_feature_name=efel_feature_name,
recording_names={"": "%s.soma.v" % protocol.name},
stim_start=stim_start,
stim_end=stim_end,
exp_mean=mean,
exp_std=0.1 * mean,
)
objective = ephys.objectives.SingletonObjective(feature_name, feature)
objectives.append(objective)
score_calc = ephys.objectivescalculators.ObjectivesCalculator(objectives)
print([(p.name, p.frozen) for p in cell.params.values()])
# cell evaluator
cell_evaluator = ephys.evaluators.CellEvaluator(
cell_model=cell,
param_names=[p.name for p in cell.params.values() if not p.frozen],
fitness_protocols={threestep_protocol.name: threestep_protocol},
fitness_calculator=score_calc,
sim=nrn,
)
# opt
optimisation = bpop.optimisations.DEAPOptimisation(
evaluator=cell_evaluator,
offspring_size=NUMBER_INDIVIDUALS,
seed=333,
)
print("run opt")
final_pop, hall_of_fame, logs, hist = optimisation.run(max_ngen=NUMBER_GENERATIONS)
# best individual
best_ind = hall_of_fame[0]
print("Best individual: ", best_ind)
print("Fitness values: ", best_ind.fitness.values)
best_ind_dict = cell_evaluator.param_dict(best_ind)
print("Ind dict", best_ind_dict)
# run protocol
responses = threestep_protocol.run(
cell_model=cell,
param_values={"gmax_glia__dbbs__Leak__0_soma": 0.00029038073716554835},
sim=nrn,
isolate=False,
)
import plotly.graph_objs as go
go.Figure(
[
go.Scatter(x=resp["time"], y=resp["voltage"], name=name)
for name, resp in responses.items()
]
).write_html("gpop.html")Please try again on latest master, or pip install arborize==4.0.0b6. Reopen this issue if the problem persists :)
danilobenozzo
commented
7 months ago Traceback (most recent call last): File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/_control.py", line 150, in runFuture future.resultValue = future.callable(*future.args, future.kargs) File "/usr/lib/python3.9/runpy.py", line 268, in run_path return _run_module_code(code, init_globals, run_name, File "/usr/lib/python3.9/runpy.py", line 97, in _run_module_code _run_code(code, mod_globals, init_globals, File "/usr/lib/python3.9/runpy.py", line 87, in _run_code exec(code, run_globals) File "granule_BluePyOpt.py", line 256, in
final_pop, hall_of_fame, logs, hist = optimisation.run(max_ngen=NUMBER_GENERATIONS) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/bluepyopt/deapext/optimisations.py", line 309, in run pop, hof, log, history = algorithms.eaAlphaMuPlusLambdaCheckpoint( File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/bluepyopt/deapext/algorithms.py", line 154, in eaAlphaMuPlusLambdaCheckpoint invalid_count = _evaluate_invalid_fitness(toolbox, population) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/bluepyopt/deapext/algorithms.py", line 61, in _evaluate_invalid_fitness fitnesses = toolbox.map(toolbox.evaluate, invalid_ind) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/fallbacks.py", line 50, in wrapper return func(*args, *kwargs) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/futures.py", line 135, in map futures = _mapFuture(func, iterables) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/futures.py", line 100, in mapFuture childrenList.append(submit(callable, args)) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/fallbacks.py", line 66, in wrapper return func(args, kwargs) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/futures.py", line 310, in submit control.execQueue.append(child) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/_types.py", line 316, in append self.socket.sendFuture(sending_future) File "/home/benozzo/dbbs_lab/test_opt/venv/lib/python3.9/site-packages/scoop/_comm/scoopzmq.py", line 369, in sendFuture pickle.dumps(future, pickle.HIGHEST_PROTOCOL), AttributeError: Can't pickle local object 'bluepyopt_build..ArborizeMorphology'
the issue was solved when running the code above which I used for testing that has only 1 generation and 1 individual, but now it happens again with NUMBER_GENERATIONS > 1 to reproduce:
import logging import bluepyopt as bpop import bluepyopt.ephys as ephys #import dbbs_models from bsb.morphologies import parse_morphology_file from arborize import define_constraints #from arborize.optimizers._bluepyopt import get_bpo_cell from arborize.builders._bluepyopt import bluepyopt_build from arborize.schematics import bsb_schematic logging.getLogger().setLevel(logging.DEBUG) def create_schema(constraints, tags, file_morphology): _m = parse_morphology_file(file_morphology, tags=tags) return bsb_schematic(_m, constraints) schema = create_schema( define_constraints( { "synapse_types": { ("AMPA", "granule"): { "tau_facil": 5, "tau_rec": 8, "tau_1": 1, "gmax": 140000, "U": 0.43, }, ("NMDA", "granule"): { "tau_facil": 5, "tau_rec": 8, "tau_1": 1, "gmax": 23500, "U": 0.43, }, ("GABA", "granule"): {"U": 0.35}, }, "cable_types": { "soma": { "cable": {"Ra": 100, "cm": 2}, "ions": {"k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}}, "mechanisms": { "Leak": { "e": -60, "gmax": [0.00029038073716, 0.00029038073717], }, "Kv3_4": {"gkbar": 0.00076192450952}, "Kv4_3": {"gkbar": 0.00281496839061}, "Kir2_3": {"gkbar": 0.00074725514702}, "Ca": {"gcabar": 0.00060938071784}, "Kv1_1": {"gbar": 0.00569738264555}, "Kv1_5": {"gKur": 0.00083407556714}, "Kv2_2": {"gKv2_2bar": 1.203410852e-05}, ("cdp5", "CR"): {}, }, }, "dendrites": { "cable": {"Ra": 100, "cm": 2.5}, "ions": {"k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}}, "mechanisms": { "Leak": {"e": -60, "gmax": 0.00025029700737}, ("Leak", "GABA"): {}, "Ca": {"gcabar": 0.00500128008459}, "Kca1_1": {"gbar": 0.01001807454651}, "Kv1_1": {"gbar": 0.00381819207934}, ("cdp5", "CR"): {}, }, }, "axon": {"cable": {}, "ions": {}, "mechanisms": {}}, "ascending_axon": { "cable": {"Ra": 100, "cm": 1}, "ions": { "na": {"rev_pot": 87.39}, "k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}, }, "mechanisms": { ("Na", "granule_cell"): {"gnabar": 0.02630163681502}, "Kv3_4": {"gkbar": 0.00237386061632}, "Leak": {"e": -60, "gmax": 9.364092125e-05}, "Ca": {"gcabar": 0.00068197420273}, ("cdp5", "CR"): {}, }, }, "parallel_fiber": { "cable": {"Ra": 100, "cm": 1}, "ions": { "na": {"rev_pot": 87.39}, "k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}, }, "mechanisms": { ("Na", "granule_cell"): {"gnabar": 0.01771848449261}, "Kv3_4": {"gkbar": 0.00817568047037}, "Leak": {"e": -60, "gmax": 3.5301616e-07}, "Ca": {"gcabar": 0.0002085683353}, ("cdp5", "CR"): {}, }, }, "axon_initial_segment": { "cable": {"Ra": 100, "cm": 1}, "ions": { "na": {"rev_pot": 87.39}, "k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}, }, "mechanisms": { ("Na", "granule_cell_FHF"): {"gnabar": 1.28725006737226}, "Kv3_4": {"gkbar": 0.00649595340654}, "Leak": {"e": -60, "gmax": 0.00029276697557}, "Ca": {"gcabar": 0.00031198539472}, "Km": {"gkbar": 0.00056671971737}, ("cdp5", "CR"): {}, }, }, "axon_hillock": { "cable": {"Ra": 100, "cm": 2}, "ions": { "na": {"rev_pot": 87.39}, "k": {"rev_pot": -80.993}, "ca": {"rev_pot": 137.5}, }, "mechanisms": { "Leak": {"e": -60, "gmax": 0.0003695818972}, ("Na", "granule_cell_FHF"): {"gnabar": 0.00928805851462}, "Kv3_4": {"gkbar": 0.02037346310915}, "Ca": {"gcabar": 0.00057726155447}, ("cdp5", "CR"): {}, }, }, }, }, use_defaults=True, ), { 16: ["axon", "axon_hillock"], 17: ["axon", "axon_initial_segment"], 18: ["axon", "ascending_axon"], 19: ["axon", "parallel_fiber"], }, r"/home/benozzo/dbbs_lab/repos/models/dbbs_models/morphologies/GranuleCell.swc", ) #cell = get_bpo_cell(schema) cell = bluepyopt_build(schema) soma_loc = ephys.locations.NrnSeclistCompLocation( name="soma", seclist_name="soma", sec_index=0, comp_x=0.5 ) NUMBER_INDIVIDUALS = 2 # Number of individuals in offspring NUMBER_GENERATIONS = 1 # Maximum number of generations sweep_protocols = [] for protocol_name, amplitude in [ # ("step1", 0.01), ("step2", 0.016), # ("step3", 0.022), ]: stim = ephys.stimuli.NrnSquarePulse( step_amplitude=amplitude, step_delay=100, step_duration=100, location=soma_loc, total_duration=200, ) rec = ephys.recordings.CompRecording( name="%s.soma.v" % protocol_name, location=soma_loc, variable="v" ) protocol = ephys.protocols.SweepProtocol(protocol_name, [stim], [rec]) sweep_protocols.append(protocol) threestep_protocol = ephys.protocols.SequenceProtocol( "twostep", protocols=sweep_protocols ) # NEURON sim nrn = ephys.simulators.NrnSimulator(dt=0.025, cvode_active=False) # feature of obj function efel_feature_means = { "step1": { "AP_height": 20.93, "ISI_CV": 0.261, "AHP_depth_abs_slow": -52.69, "AP_width": 0.665, "voltage_base": -68.5, "AHP_depth_abs": -59.21, "time_to_first_spike": 31.9, "adaptation_index2": 0.1062, "mean_frequency": 25, }, "step2": { "AP_height": 19.255, "ISI_CV": 0.14, "AHP_depth_abs_slow": -48.935, "AP_width": 0.695, "voltage_base": -68.77, "AHP_depth_abs": -58.3, "time_to_first_spike": 19.0, "adaptation_index2": 0.034, "mean_frequency": 40, }, "step3": { "AP_height": 17.645, "ISI_CV": 0.148, "AHP_depth_abs_slow": -32.67, "AP_width": 0.7135, "voltage_base": -69.125, "AHP_depth_abs": -57.191, "time_to_first_spike": 14.65, "adaptation_index2": 0.029, "mean_frequency": 50, }, } # obj function objectives = [] for protocol in sweep_protocols: stim_start = protocol.stimuli[0].step_delay stim_end = stim_start + protocol.stimuli[0].step_duration for efel_feature_name, mean in efel_feature_means[protocol.name].items(): feature_name = "%s.%s" % (protocol.name, efel_feature_name) feature = ephys.efeatures.eFELFeature( feature_name, efel_feature_name=efel_feature_name, recording_names={"": "%s.soma.v" % protocol.name}, stim_start=stim_start, stim_end=stim_end, exp_mean=mean, exp_std=0.1 * mean, ) objective = ephys.objectives.SingletonObjective(feature_name, feature) objectives.append(objective) score_calc = ephys.objectivescalculators.ObjectivesCalculator(objectives) print([(p.name, p.frozen) for p in cell.params.values()]) # cell evaluator cell_evaluator = ephys.evaluators.CellEvaluator( cell_model=cell, param_names=[p.name for p in cell.params.values() if not p.frozen], fitness_protocols={threestep_protocol.name: threestep_protocol}, fitness_calculator=score_calc, sim=nrn, ) # opt optimisation = bpop.optimisations.DEAPOptimisation( evaluator=cell_evaluator, offspring_size=NUMBER_INDIVIDUALS, seed=333, ) print("run opt") final_pop, hall_of_fame, logs, hist = optimisation.run(max_ngen=NUMBER_GENERATIONS) # best individual best_ind = hall_of_fame[0] print("Best individual: ", best_ind) print("Fitness values: ", best_ind.fitness.values) best_ind_dict = cell_evaluator.param_dict(best_ind) print("Ind dict", best_ind_dict) # run protocol responses = threestep_protocol.run( cell_model=cell, param_values={"gmax_glia__dbbs__Leak__0_soma": 0.00029038073716554835}, sim=nrn, isolate=False, ) import plotly.graph_objs as go go.Figure( [ go.Scatter(x=resp["time"], y=resp["voltage"], name=name) for name, resp in responses.items() ] ).write_html("gpop.html")
when running the optimization moving back to what was in /origin/feature/optimization works