def launch_nk(params,debug=False,device=0,log=False):
""" Launch Neurokernel
"""
if log:
screen = True
logger = setup_logger(file_name=params['name'] +'.log', screen=screen)
man = Manager()
man.add(LPU,
params['name'],
params['dt'],
params['n_dict'],
params['s_dict'],
input_file=params['input_file'],
output_file=params['output_file'],
device=device,
debug=debug,
components=params['components'])
man.spawn()
man.start(int(params['steps']))
man.wait()
def run_lif(payload):
"""
Run an example lif model through the nk_server json interface
"""
print payload
payload = json.loads(payload)
params = payload['params']
sim_input = payload['input']
data_dir = 'data/'+params['sim_uid'] + '/' + params['sim_exp'] +'/'
try:
os.stat('data/'+params['sim_uid']+'/')
except:
os.mkdir('data/'+params['sim_uid']+'/')
try:
os.stat(data_dir)
except:
os.mkdir(data_dir)
dt = params['sim_dt']
Nt = params['sim_steps']
dur = Nt/dt
G = nx.DiGraph() # or nx.MultiDiGraph()
G.add_nodes_from([0])
G.node[0] = {
'model': 'LeakyIAF_rest',
'name': 'neuron_0',
'extern': True, # indicates whether the neuron can receive an external input signal
'public': True, # indicates whether the neuron can emit output to other LPUs
'spiking': True, # indicates whether the neuron outputs spikes or a membrane voltage
'selector': '/a[0]', # every public neuron must have a selector
'V': params['par_V'], # initial membrane voltage
'Vr': params['par_Vr'], # reset voltage ## The same as the implicit resting potential
'Vt': params['par_Vt'], # spike threshold
'R': params['par_R'], # membrane resistance
'C': params['par_C'], # membrane capacitance
'Er': params['par_rest'] # membrane capacitance
}
nx.write_gexf(G, data_dir +'lif_graph.gexf.gz')
N_neurons = G.number_of_nodes()
if sim_input == 'Default':
t = np.arange(0, params['sim_dt']*params['sim_steps'], params['sim_dt'])
I = np.zeros((params['sim_steps'], N_neurons), dtype=np.double)
I[t>0.2] = 1e-9
I[t>0.4] = 2e-9
I[t>0.8] = 2.5e-9
with h5py.File(data_dir + 'lif_input.h5', 'w') as f:
f.create_dataset('array', (Nt, N_neurons),
dtype=np.double,
data=I)
else:
print 'loading non-default inputs (WIP)'
with h5py.File(data_dir + 'lif_input.h5', 'w') as f:
f.create_dataset('array', (Nt, N_neurons),
dtype=np.double,
data=sim_input)
parser = argparse.ArgumentParser()
parser.add_argument('--debug', default=False,
dest='debug', action='store_true',
help='Write connectivity structures and inter-LPU routed data in debug folder')
parser.add_argument('-l', '--log', default='none', type=str,
help='Log output to screen [file, screen, both, or none; default:none]')
parser.add_argument('-s', '--steps', default=params['sim_steps'], type=int,
help='Number of steps [default: %s]' % params['sim_steps'])
args = parser.parse_args()
file_name = None
screen = False
if args.log.lower() in ['file', 'both']:
file_name = 'lif.log'
if args.log.lower() in ['screen', 'both']:
screen = True
logger = setup_logger(file_name=file_name, screen=screen)
man = Manager()
(n_dict, s_dict) = LPU.lpu_parser(data_dir+'lif_graph.gexf.gz')
if params['sim_output'] != 'spike':
args.debug = True
man.add(LPU, 'lif', dt, n_dict, s_dict,
input_file=data_dir +'lif_input.h5',
output_file=data_dir +'lif_output.h5',
device=0, debug=True)
man.spawn()
man.start(steps=params['sim_steps'])
man.wait()
if params['sim_output'] == 'spike':
with h5py.File(data_dir +'lif_output_spike.h5') as f:
data = np.array(f['array']).T.tolist()
else:
######## BUG: Needs to output debug to data folder
with h5py.File('./lif_V.h5') as f:
data = np.array(f['array']).T.tolist()
return data
def setUp(self):
self.man = Manager()
class test_core_gpu(TestCase):
def setUp(self):
self.man = Manager()
def test_transmit_spikes_one_to_one(self):
m1_sel_in_gpot = Selector('')
m1_sel_out_gpot = Selector('')
m1_sel_in_spike = Selector('')
m1_sel_out_spike = Selector('/m1/out/spike[0:4]')
m1_sel, m1_sel_in, m1_sel_out, m1_sel_gpot, m1_sel_spike = \
make_sels(m1_sel_in_gpot, m1_sel_out_gpot, m1_sel_in_spike, m1_sel_out_spike)
N1_gpot = SelectorMethods.count_ports(m1_sel_gpot)
N1_spike = SelectorMethods.count_ports(m1_sel_spike)
m2_sel_in_gpot = Selector('')
m2_sel_out_gpot = Selector('')
m2_sel_in_spike = Selector('/m2/in/spike[0:4]')
m2_sel_out_spike = Selector('')
m2_sel, m2_sel_in, m2_sel_out, m2_sel_gpot, m2_sel_spike = \
make_sels(m2_sel_in_gpot, m2_sel_out_gpot, m2_sel_in_spike, m2_sel_out_spike)
N2_gpot = SelectorMethods.count_ports(m2_sel_gpot)
N2_spike = SelectorMethods.count_ports(m2_sel_spike)
m1_id = 'm1'
self.man.add(MyModule1, m1_id,
m1_sel, m1_sel_in, m1_sel_out,
m1_sel_gpot, m1_sel_spike,
np.zeros(N1_gpot, dtype=np.double),
np.zeros(N1_spike, dtype=int),
device=0, debug=debug, out_spike_data=[0, 0, 1, 1])
f, out_file_name = tempfile.mkstemp()
os.close(f)
m2_id = 'm2'
self.man.add(MyModule2, m2_id,
m2_sel, m2_sel_in, m2_sel_out,
m2_sel_gpot, m2_sel_spike,
np.zeros(N2_gpot, dtype=np.double),
np.zeros(N2_spike, dtype=int),
device=1, debug=debug, out_file_name=out_file_name)
pat12 = Pattern(m1_sel, m2_sel)
pat12.interface[m1_sel_out_gpot] = [0, 'in', 'gpot']
pat12.interface[m1_sel_in_gpot] = [0, 'out', 'gpot']
pat12.interface[m1_sel_out_spike] = [0, 'in', 'spike']
pat12.interface[m1_sel_in_spike] = [0, 'out', 'spike']
pat12.interface[m2_sel_in_gpot] = [1, 'out', 'gpot']
pat12.interface[m2_sel_out_gpot] = [1, 'in', 'gpot']
pat12.interface[m2_sel_in_spike] = [1, 'out', 'spike']
pat12.interface[m2_sel_out_spike] = [1, 'in', 'spike']
pat12['/m1/out/spike[0]', '/m2/in/spike[0]'] = 1
pat12['/m1/out/spike[1]', '/m2/in/spike[1]'] = 1
pat12['/m1/out/spike[2]', '/m2/in/spike[2]'] = 1
pat12['/m1/out/spike[3]', '/m2/in/spike[3]'] = 1
self.man.connect(m1_id, m2_id, pat12, 0, 1)
# Run emulation for 2 steps:
self.man.spawn()
self.man.start(2)
self.man.wait()
# Get output of m2:
with open(out_file_name, 'r') as f:
output = pickle.load(f)
os.remove(out_file_name)
self.assertSequenceEqual(list(output), [0, 0, 1, 1])
def test_transmit_spikes_one_to_many(self):
m1_sel_in_gpot = Selector('')
m1_sel_out_gpot = Selector('')
m1_sel_in_spike = Selector('')
m1_sel_out_spike = Selector('/m1/out/spike[0:4]')
m1_sel, m1_sel_in, m1_sel_out, m1_sel_gpot, m1_sel_spike = \
make_sels(m1_sel_in_gpot, m1_sel_out_gpot, m1_sel_in_spike, m1_sel_out_spike)
N1_gpot = SelectorMethods.count_ports(m1_sel_gpot)
N1_spike = SelectorMethods.count_ports(m1_sel_spike)
m2_sel_in_gpot = Selector('')
m2_sel_out_gpot = Selector('')
m2_sel_in_spike = Selector('/m2/in/spike[0:4]')
m2_sel_out_spike = Selector('')
m2_sel, m2_sel_in, m2_sel_out, m2_sel_gpot, m2_sel_spike = \
make_sels(m2_sel_in_gpot, m2_sel_out_gpot, m2_sel_in_spike, m2_sel_out_spike)
N2_gpot = SelectorMethods.count_ports(m2_sel_gpot)
N2_spike = SelectorMethods.count_ports(m2_sel_spike)
m1_id = 'm1'
self.man.add(MyModule1, m1_id,
m1_sel, m1_sel_in, m1_sel_out,
m1_sel_gpot, m1_sel_spike,
np.zeros(N1_gpot, dtype=np.double),
np.zeros(N1_spike, dtype=int),
device=0, debug=debug, out_spike_data=[1, 0, 0, 0])
f, out_file_name = tempfile.mkstemp()
os.close(f)
m2_id = 'm2'
self.man.add(MyModule2, m2_id,
m2_sel, m2_sel_in, m2_sel_out,
m2_sel_gpot, m2_sel_spike,
np.zeros(N2_gpot, dtype=np.double),
np.zeros(N2_spike, dtype=int),
device=1, debug=debug, out_file_name=out_file_name)
pat12 = Pattern(m1_sel, m2_sel)
pat12.interface[m1_sel_out_gpot] = [0, 'in', 'gpot']
pat12.interface[m1_sel_in_gpot] = [0, 'out', 'gpot']
pat12.interface[m1_sel_out_spike] = [0, 'in', 'spike']
pat12.interface[m1_sel_in_spike] = [0, 'out', 'spike']
pat12.interface[m2_sel_in_gpot] = [1, 'out', 'gpot']
pat12.interface[m2_sel_out_gpot] = [1, 'in', 'gpot']
pat12.interface[m2_sel_in_spike] = [1, 'out', 'spike']
pat12.interface[m2_sel_out_spike] = [1, 'in', 'spike']
pat12['/m1/out/spike[0]', '/m2/in/spike[0]'] = 1
pat12['/m1/out/spike[0]', '/m2/in/spike[1]'] = 1
pat12['/m1/out/spike[0]', '/m2/in/spike[2]'] = 1
pat12['/m1/out/spike[0]', '/m2/in/spike[3]'] = 1
self.man.connect(m1_id, m2_id, pat12, 0, 1)
# Run emulation for 2 steps:
self.man.spawn()
self.man.start(2)
self.man.wait()
# Get output of m2:
with open(out_file_name, 'r') as f:
output = pickle.load(f)
os.remove(out_file_name)
self.assertSequenceEqual(list(output), [1, 1, 1, 1])
def compile(self, duration, dt=None, steps=None, in_list=None,
record=('V', 'spike_state', 'I'), extra_comps=None,
input_filename='neuroballad_temp_model_input.h5',
output_filename='neuroballad_temp_model_output.h5',
graph_filename='neuroballand_temp_graph.gexf.gz',
device=0, sample_interval=1):
if dt is not None:
if steps is not None:
assert dt*steps == duration, 'dt*step != duration'
else:
steps = int(duration/dt)
t = np.linspace(0, duration, steps)
else:
if steps is not None:
t = np.linspace(0, duration, steps)
dt = t[1] - t[0]
else:
raise ValueError('dt and step cannot both be None')
self.config = self.config._replace(duration=duration,
steps=steps,
dt=dt,
t=t,
device=device)
# compile inputs
if in_list is None:
in_list = self._inputs
uids = []
for i in in_list:
uids.append(self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
input_vars = []
for i in in_list:
if isinstance(i.var, list):
for j in i.var:
input_vars.append(j)
else:
input_vars.append(i.var)
input_vars = list(set(input_vars))
uids = np.array(list(set(uids)), dtype='S')
Is = {}
Inodes = {}
for i in input_vars:
Inodes[i] = []
for i in in_list:
in_name = self.encode_name(str(i.node_id),
experiment_name=i.experiment_name)
if in_name in list(self.G.nodes(data=False)):
pass
else:
raise ValueError(
'Input node {} not found in Circuit.'.format(in_name))
if isinstance(i.var, list):
for j in i.var:
Inodes[j].append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
else:
Inodes[i.var].append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
for i in input_vars:
Inodes[i] = np.array(list(set(Inodes[i])), dtype='S')
for i in input_vars:
Is[i] = np.zeros((self.config.steps, len(Inodes[i])),
dtype=self.dtype)
for i in in_list:
if isinstance(i.var, list):
for j in i.var:
Is[j] = i.add(self, Inodes[j], Is[j], t, var=j)
else:
Is[i.var] = i.add(self, Inodes[i.var], Is[i.var], t, var=i.var)
with h5py.File(input_filename, 'w') as f:
for i in input_vars:
# print(i + '/uids')
i_nodes = Inodes[i]
"""
try:
i_nodes = [i.decode('ascii') for i in i_nodes]
except:
pass
i_nodes = [self.encode_name(i) for i in i_nodes]
"""
i_nodes = np.array(i_nodes, dtype='S')
f.create_dataset(i + '/uids', data=i_nodes)
f.create_dataset(i + '/data', (self.config.steps, len(Inodes[i])),
dtype=self.dtype,
data=Is[i])
if graph_filename is not None:
nx.write_gexf(self.G, graph_filename)
from neurokernel.core_gpu import Manager
from neurokernel.LPU.LPU import LPU
import neurokernel.mpi_relaunch
from neurokernel.LPU.InputProcessors.FileInputProcessor import \
FileInputProcessor
from neurokernel.LPU.OutputProcessors.FileOutputProcessor import \
FileOutputProcessor
input_processor = FileInputProcessor(input_filename)
(comp_dict, conns) = LPU.graph_to_dicts(self.G)
output_processor = FileOutputProcessor([(i, None) for i in list(record)],
output_filename,
sample_interval=sample_interval)
self.manager = Manager()
self.manager.add(LPU, self.experiment_name, self.config.dt,
comp_dict, conns,
device=self.config.device,
input_processors=[input_processor],
output_processors=[output_processor],
debug=False,
extra_comps=extra_comps if extra_comps is not None else [])
class Circuit(object):
"""
Create a Neuroballad circuit.
Basic Example
--------
>>> from neuroballad import * #Import Neuroballad
>>> C.add([0, 2, 4], HodgkinHuxley()) #Create three neurons
>>> C.add([1, 3, 5], AlphaSynapse()) #Create three synapses
>>> C.join([[0,1],[1,2],[2,3],[3,4],[4,5],[5,0]]) #Join nodes together
>>> C_in_a = InIStep(0, 40., 0.25, 0.50) #Create current input for node 0
>>> C_in_b = InIStep(2, 40., 0.50, 0.75) #Create current input for node 2
>>> C_in_c = InIStep(4, 40., 0.75, 0.50) #Create current input for node 4
>>> C.sim(1., 1e-4, [C_in_a, C_in_b, C_in_c]) #Use the inputs and simulate
"""
default_tags = {'species': 'None',
'hemisphere': 'None',
'neuropil': 'None',
'circuit': 'None',
'from': 'None',
'to': 'None',
'neurotransmitter': 'None',
'experiment_id': 'None',
'rid': 'None',
'type': 'neuron'}
def __init__(self, name='', dtype=np.float64, experiment_name=''):
self.G = nx.MultiDiGraph() # Neurokernel graph definition
self.config = SimConfig(duration=None, steps=None,
dt=1e-4, t=None, device=0)
self.node_ids = [] # Graph ID's
self.tracked_variables = [] # Observable variables in circuit
self._inputs = None # input nodes
# self._outputs = None # output nodes
# self.experiment_config = []
self.experiment_name = experiment_name
self.dtype = dtype
# self.ICs = []
self.name = name
self.manager = None # LPU Manager
self.logger = None # logger
def set_experiment(self, experiment_name):
self.experiment_name = experiment_name
Gc = copy.deepcopy(self.G)
mapping = {}
for i in self.G.nodes():
ii = self.json_to_tags(i)
ii['experiment_name'] = experiment_name
mapping[i] = self.tags_to_json(ii)
Gc = nx.relabel_nodes(Gc, mapping)
self.G = Gc
for i, val in self.G.nodes(data=True):
val['name'] = i
for i in range(len(self.node_ids)):
self.node_ids[i][1] = experiment_name
def get_new_id(self):
"""Generate new ID
"""
if self.node_ids == []:
return '1'
else:
return str(len(self.node_ids)+1)
@property
def nodes(self):
return self.G.nodes
def copy(self):
'''Return Copy of Circuit Instance'''
return copy.deepcopy(self)
def merge(self, C):
'''Merge circuit `C` with current circuit'''
self.G = nx.compose(self.G, C.G)
self.node_ids += C.node_ids
def find_neurons(self, **tags_tofind):
tags_tofind['type'] = 'neuron'
return self.filter_nodes_by_tags(tags_tofind)
def find_synapses(self, **tags_tofind):
tags_tofind['type'] = 'synapse'
return self.filter_nodes_by_tags(tags_tofind)
def find_ports(self, **tags_tofind):
tags_tofind['type'] = 'port'
return self.filter_nodes_by_tags(tags_tofind)
def filter_nodes_by_tags(self, tags_tofind):
output = []
for i, val in self.G.nodes(data=True):
skip = False
for j in tags_tofind.keys():
if j not in val:
skip = True
else:
if tags_tofind[j] in val[j]:
pass
else:
skip = True
if not skip:
output.append(i)
return output
def tags_to_json(self, tags):
"""
Turns the tags dictionary to a JSON string.
"""
return json.dumps(tags)
def json_to_tags(self, tags_str):
"""
Turns a tags JSON string to the dictionary.
"""
return json.loads(tags_str)
def encode_name(self, i, experiment_name=None):
'''Encode node id into json format
Example
-------
>>> a = C.encode_name(0, experiment_name='test')
>>> a
'{"name": "0", "experiment_name": "test"}'
'''
i = str(i)
try:
i = i.decode('ascii')
except Exception as e:
pass
# TODO: use raise ValueError('ASCII decode failed for {}, error {}'.format(i, e))
if experiment_name is None:
experiment_name = self.experiment_name
name_dict = {'name': str(i), 'exp': experiment_name}
name = self.tags_to_json(name_dict)
return name
def add(self, name, neuron):
"""
Loops through a list of ID inputs and adds corresponding components
of a specific type to the circuit.
Example
--------
>>> C.add([1, 2, 3], HodgkinHuxley())
"""
if isinstance(neuron, Input):
self._inputs.append(neuron.addToExperiment)
else:
for i in name:
if i in self.node_ids:
raise ValueError(
'Don''t use the same ID for multiple neurons!')
for i in name:
neuron.nadd(self, i, self.experiment_name, self.default_tags)
self.node_ids.append([str(i), self.experiment_name])
def add_cluster(self, number, neuron, name=None):
"""
Creates a number of components of a specific type and returns their
ID's.
Example
--------
>>> id_list = C.add_cluster(256, HodgkinHuxley())
"""
cluster_inds = []
for i in range(number):
i_toadd = self.get_new_id()
_id = '{}-{}'.format(name, i_toadd)
neuron.nadd(self, _id, self.experiment_name, self.default_tags)
self.node_ids.append(_id)
cluster_inds.append(_id)
return cluster_inds
def dense_connect_variable(self, in_array_a, in_array_b, neuron,
delay=0.0, variable='', tag=0, debug=False):
"""
Densely connects two arrays of circuit ID's, creating a layer of unique
components of a specified type in between.
Example
--------
>>> C.dense_join_variable(cluster_a, cluster_b, AlphaSynapse())
"""
for i in in_array_a:
for j in in_array_b:
i_toadd = self.get_new_id()
if debug:
print('Added neuron ID: ' + str(i_toadd))
neuron.nadd(self, i_toadd, self.experiment_name,
self.default_tags)
self.node_ids.append(i_toadd)
self.join([[i, i_toadd], [i_toadd, j]],
delay=delay, variable=variable, tag=tag)
def dense_connect(self, in_array_a, in_array_b, delay=0.0):
"""Densely connect clusters
Densely connects two arrays of circuit ID's.
Example
--------
>>> C.dense_connect_variable(cluster_a, cluster_b)
"""
for i in in_array_a:
for j in in_array_b:
self.join([[i, j]], delay=delay)
def dense_join_variable(self, in_array_a, in_array_b, in_array_c, delay=0.0, variable=None):
"""Densely connect clusters variable intermediary elements
Densely connects two arrays of circuit ID's, using a third array as the
matrix of components that connects the two.
TODO
----
1. Currently only support scalar delay, add component-dependent
delay
Example
--------
>>> C.dense_join_variable(cluster_a, cluster_b, cluster_c)
"""
if np.isscalar(delay):
delay = [delay]*2
if isinstance(variable, str):
variable = [variable]*2
k = 0
in_array_c = in_array_c.flatten()
for i in in_array_a:
for j in in_array_b:
if variable is not None:
self.join([i, in_array_c[k]], delay=delay[0],
variable=variable[0])
self.join([in_array_c[k], j], delay=delay[1],
variable=variable[1])
else:
self.join([i, in_array_c[k]], delay=delay[0])
self.join([in_array_c[k], j], delay=delay[1])
k += 1
def join(self, in_array, delay=0.0, variable=None, tag=0):
"""
Processes an edge list and adds the edges to the circuit.
Example
--------
>>> C.add([0, 2, 4], HodgkinHuxley()) #Create three neurons
>>> C.add([1, 3, 5], AlphaSynapse()) #Create three synapses
>>> C.join([[0,1],[1,2],[2,3],[3,4],[4,5],[5,0]])
"""
in_array = np.array(in_array)
# print(in_array)
for i in range(in_array.shape[0]):
if variable is None:
self.G.add_edge(self.encode_name(str(in_array[i, 0])),
self.encode_name(str(in_array[i, 1])),
delay=delay,
tag=tag)
else:
self.G.add_edge(self.encode_name(str(in_array[i, 0])),
self.encode_name(str(in_array[i, 1])),
delay=delay,
variable=variable,
tag=tag)
def fit(self, inputs):
"""
Attempts to find parameters to fit a certain curve to the output.
Not implemented at this time.
"""
raise NotImplementedError
def load_last(self, file_name='neuroballad_temp_model.gexf.gz'):
"""
Loads the latest executed circuit in the directory.
Example
--------
>>> C.load_last()
"""
self.G = nx.read_gexf(file_name)
self.node_ids = []
for i in self.G.nodes():
self.node_ids.append(i) # FIX
def save(self, file_name='neuroballad_temp_model.gexf.gz'):
"""
Saves the current circuit to a file.
Example
--------
>>> C.save(file_name = 'my_circuit.gexf.gz')
"""
nx.write_gexf(self.G, file_name)
def compile(self, duration, dt=None, steps=None, in_list=None,
record=('V', 'spike_state', 'I'), extra_comps=None,
input_filename='neuroballad_temp_model_input.h5',
output_filename='neuroballad_temp_model_output.h5',
graph_filename='neuroballand_temp_graph.gexf.gz',
device=0, sample_interval=1):
if dt is not None:
if steps is not None:
assert dt*steps == duration, 'dt*step != duration'
else:
steps = int(duration/dt)
t = np.linspace(0, duration, steps)
else:
if steps is not None:
t = np.linspace(0, duration, steps)
dt = t[1] - t[0]
else:
raise ValueError('dt and step cannot both be None')
self.config = self.config._replace(duration=duration,
steps=steps,
dt=dt,
t=t,
device=device)
# compile inputs
if in_list is None:
in_list = self._inputs
uids = []
for i in in_list:
uids.append(self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
input_vars = []
for i in in_list:
if isinstance(i.var, list):
for j in i.var:
input_vars.append(j)
else:
input_vars.append(i.var)
input_vars = list(set(input_vars))
uids = np.array(list(set(uids)), dtype='S')
Is = {}
Inodes = {}
for i in input_vars:
Inodes[i] = []
for i in in_list:
in_name = self.encode_name(str(i.node_id),
experiment_name=i.experiment_name)
if in_name in list(self.G.nodes(data=False)):
pass
else:
raise ValueError(
'Input node {} not found in Circuit.'.format(in_name))
if isinstance(i.var, list):
for j in i.var:
Inodes[j].append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
else:
Inodes[i.var].append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
for i in input_vars:
Inodes[i] = np.array(list(set(Inodes[i])), dtype='S')
for i in input_vars:
Is[i] = np.zeros((self.config.steps, len(Inodes[i])),
dtype=self.dtype)
for i in in_list:
if isinstance(i.var, list):
for j in i.var:
Is[j] = i.add(self, Inodes[j], Is[j], t, var=j)
else:
Is[i.var] = i.add(self, Inodes[i.var], Is[i.var], t, var=i.var)
with h5py.File(input_filename, 'w') as f:
for i in input_vars:
# print(i + '/uids')
i_nodes = Inodes[i]
"""
try:
i_nodes = [i.decode('ascii') for i in i_nodes]
except:
pass
i_nodes = [self.encode_name(i) for i in i_nodes]
"""
i_nodes = np.array(i_nodes, dtype='S')
f.create_dataset(i + '/uids', data=i_nodes)
f.create_dataset(i + '/data', (self.config.steps, len(Inodes[i])),
dtype=self.dtype,
data=Is[i])
if graph_filename is not None:
nx.write_gexf(self.G, graph_filename)
from neurokernel.core_gpu import Manager
from neurokernel.LPU.LPU import LPU
import neurokernel.mpi_relaunch
from neurokernel.LPU.InputProcessors.FileInputProcessor import \
FileInputProcessor
from neurokernel.LPU.OutputProcessors.FileOutputProcessor import \
FileOutputProcessor
input_processor = FileInputProcessor(input_filename)
(comp_dict, conns) = LPU.graph_to_dicts(self.G)
output_processor = FileOutputProcessor([(i, None) for i in list(record)],
output_filename,
sample_interval=sample_interval)
self.manager = Manager()
self.manager.add(LPU, self.experiment_name, self.config.dt,
comp_dict, conns,
device=self.config.device,
input_processors=[input_processor],
output_processors=[output_processor],
debug=False,
extra_comps=extra_comps if extra_comps is not None else [])
# self.input.status = 'pre_run'
# self.output.status = 'pre_run'
def sim(self, duration, dt, steps=None, in_list=None,
record=('V', 'spike_state', 'I'), log=None,
device=0, sample_interval=1,
input_filename='neuroballad_temp_model_input.h5',
output_filename='neuroballad_temp_model_output.h5',
graph_filename='neuroballand_temp_graph.gexf.gz',
log_filename='neuroballand_temp_log.log',
extra_comps=None, preamble=[], args=[]):
"""
Simulates the circuit for a set amount of time, with a fixed temporal
step size and a list of inputs.
TODO
----
1. use preamble and args for slurm
Example
--------
>>> C.sim(1., 1e-4, InIStep(0, 10., 1., 2.))
"""
from neurokernel.tools.logging import setup_logger
if log is not None:
screen = False
file_name = None
if log.lower() in ['file', 'both']:
file_name = log_filename
if log.lower() in ['screen', 'both']:
screen = True
self.logger = setup_logger(file_name=file_name, screen=screen)
self.compile(duration, dt, steps=steps,
in_list=in_list, record=record, extra_comps=extra_comps,
input_filename=input_filename, output_filename=output_filename,
graph_filename=graph_filename,
device=device, sample_interval=sample_interval)
self.manager.spawn()
self.manager.start(self.config.steps)
self.manager.wait()
def collect(self):
data = {'in': {}, 'out': {}}
uids = {'in': {}, 'out': {}}
time.sleep(1.)
with h5py.File('neuroballad_temp_model_input.h5', 'r') as f:
for k in f.keys():
data['in'][k] = f[k]['data']
uids['in'][k] = f[k]['uids'].astype(str)
with h5py.File('neuroballad_temp_model_output.h5', 'r') as f:
for k in f.keys():
if k != 'metadata':
data['out'][k] = f[k]['data'][()]
uids['out'][k] = f[k]['uids'][()].astype(str)
return uids, data
def collect_results(self):
"""
Collects the latest results from the executor. Useful when loading
a set of results after execution.
Example
--------
>>> C.collect_results()
"""
import neurokernel.LPU.utils.visualizer as vis
self.Viz = vis.visualizer()
self.Viz.add_LPU('neuroballad_temp_model_output.h5',
gexf_file='neuroballad_temp_model.gexf.gz', LPU='lpu')
def visualize_circuit(self, prog='dot', splines='line',
filename='neuroballad_temp_circuit.svg'):
return visualize_circuit(self,
prog=prog,
splines=splines,
filename=filename)
def visualize_video(self, name, config={}, visualization_variable='V',
out_name='test.avi'):
visualize_video(self,
name=name,
config=config,
visualization_variable=visualization_variable,
out_name=out_name)
def emulate(n_lpu, n_spike, n_gpot, steps):
"""
Benchmark inter-LPU communication throughput.
Each LPU is configured to use a different local GPU.
Parameters
----------
n_lpu : int
Number of LPUs. Must be at least 2 and no greater than the number of
local GPUs.
n_spike : int
Total number of input and output spiking ports any
single LPU exposes to any other LPU. Each LPU will therefore
have 2*n_spike*(n_lpu-1) total spiking ports.
n_gpot : int
Total number of input and output graded potential ports any
single LPU exposes to any other LPU. Each LPU will therefore
have 2*n_gpot*(n_lpu-1) total graded potential ports.
steps : int
Number of steps to execute.
Returns
-------
average_throughput, total_throughput : float
Average per-step and total received data throughput in bytes/seconds.
exec_time : float
Execution time in seconds.
"""
# Time everything starting with manager initialization:
start_all = time.time()
# Check whether a sufficient number of GPUs are available:
drv.init()
if n_lpu > drv.Device.count():
raise RuntimeError('insufficient number of available GPUs.')
# Set up manager:
man = Manager()
# Generate selectors for configuring modules and patterns:
mod_sels, pat_sels = gen_sels(n_lpu, n_spike, n_gpot)
# Set up modules:
for i in xrange(n_lpu):
lpu_i = 'lpu%s' % i
sel, sel_in, sel_out, sel_gpot, sel_spike = mod_sels[lpu_i]
sel = Selector.union(sel_in, sel_out, sel_gpot, sel_spike)
man.add(MyModule, lpu_i, sel, sel_in, sel_out,
sel_gpot, sel_spike,
None, None, ['interface', 'io', 'type'],
CTRL_TAG, GPOT_TAG, SPIKE_TAG,
device=i, time_sync=True)
# Set up connections between module pairs:
for i, j in itertools.combinations(xrange(n_lpu), 2):
lpu_i = 'lpu%s' % i
lpu_j = 'lpu%s' % j
sel_from, sel_to, sel_in_i, sel_out_i, sel_gpot_i, sel_spike_i, \
sel_in_j, sel_out_j, sel_gpot_j, sel_spike_j = pat_sels[(lpu_i, lpu_j)]
pat = Pattern.from_concat(sel_from, sel_to,
from_sel=sel_from, to_sel=sel_to, data=1)
pat.interface[sel_in_i, 'interface', 'io'] = [0, 'in']
pat.interface[sel_out_i, 'interface', 'io'] = [0, 'out']
pat.interface[sel_gpot_i, 'interface', 'type'] = [0, 'gpot']
pat.interface[sel_spike_i, 'interface', 'type'] = [0, 'spike']
pat.interface[sel_in_j, 'interface', 'io'] = [1, 'in']
pat.interface[sel_out_j, 'interface', 'io'] = [1, 'out']
pat.interface[sel_gpot_j, 'interface', 'type'] = [1, 'gpot']
pat.interface[sel_spike_j, 'interface', 'type'] = [1, 'spike']
man.connect(lpu_i, lpu_j, pat, 0, 1, compat_check=False)
man.spawn()
start_main = time.time()
man.start(steps)
man.wait()
stop_main = time.time()
return man.average_step_sync_time, (time.time()-start_all), \
(stop_main-start_main), (man.stop_time-man.start_time)
def compile(self,
duration,
dt=None,
steps=None,
in_list=None,
record=('V', 'spike_state', 'I'),
extra_comps=None,
input_filename='neuroballad_temp_model_input.h5',
output_filename='neuroballad_temp_model_output.h5',
graph_filename='neuroballad_temp_graph.gexf.gz',
device=0,
sample_interval=1,
execute_in_same_thread=True):
"""
Compiles a neuroballad circuit before execution.
# Arguments
duration (float): Simulation duration.
dt (float): Time step size.
steps (int): Number of steps to use in simulation. Optional; don't use dt if provided.
in_list (list): List of inputs to use during compilation.
record (tuple): Tuple of variables to record. Defaults to ('V', 'spike_state', 'I').
extra_comps (list): List of new, custom components to include for your simulation.
input_filename (str): The .h5 file name to use for the input.
output_filename (str): The .h5 file name to use for recording the output.
graph_filename (str): Name of the graph file to save the circuit to. Uses the .gexf format.
device (int): Device to use for execution.
sample_interval (int): Sampling interval for recording simulation output.
execute_in_same_thread (bool): Whether to execute the circuit in the current thread.
"""
if dt is not None:
if steps is not None:
warnings.warn(
"Both 'steps' and 'duration' arguments were specified. 'steps' argument is ignored."
)
steps = int(duration / dt)
else:
steps = int(duration / dt)
t = np.linspace(0, duration, steps)
else:
if steps is not None:
t = np.linspace(0, duration, steps)
dt = t[1] - t[0]
else:
raise ValueError('dt and step cannot both be None')
self.config = self.config._replace(duration=duration,
steps=steps,
dt=dt,
t=t,
device=device)
run_parameters = [duration, dt]
with open('run_parameters.pickle', 'wb') as f:
pickle.dump(run_parameters, f, protocol=pickle.HIGHEST_PROTOCOL)
# Compile inputs
if in_list is None:
in_list = self._inputs
uids = []
for i in in_list:
uids.append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
input_vars = []
for i in in_list:
if isinstance(i.var, list):
for j in i.var:
input_vars.append(j)
else:
input_vars.append(i.var)
input_vars = list(set(input_vars))
uids = np.array(list(set(uids)), dtype='S')
Is = {}
Inodes = {}
for i in input_vars:
Inodes[i] = []
for i in in_list:
in_name = self.encode_name(str(i.node_id),
experiment_name=i.experiment_name)
if in_name in list(self.G.nodes(data=False)):
pass
else:
raise ValueError(
'Input node {} not found in Circuit.'.format(in_name))
if isinstance(i.var, list):
for j in i.var:
Inodes[j].append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
else:
Inodes[i.var].append(
self.encode_name(str(i.node_id),
experiment_name=i.experiment_name))
for i in input_vars:
Inodes[i] = np.array(list(set(Inodes[i])), dtype='S')
for i in input_vars:
Is[i] = np.zeros((self.config.steps, len(Inodes[i])),
dtype=self.dtype)
for i in in_list:
if isinstance(i.var, list):
for j in i.var:
Is[j] = i.add(self, Inodes[j], Is[j], t, var=j)
else:
Is[i.var] = i.add(self, Inodes[i.var], Is[i.var], t, var=i.var)
with h5py.File(input_filename, 'w') as f:
for i in input_vars:
# print(i + '/uids')
i_nodes = Inodes[i]
"""
try:
i_nodes = [i.decode('ascii') for i in i_nodes]
except:
pass
i_nodes = [self.encode_name(i) for i in i_nodes]
"""
i_nodes = np.array(i_nodes, dtype='S')
f.create_dataset(i + '/uids', data=i_nodes)
f.create_dataset(i + '/data',
(self.config.steps, len(Inodes[i])),
dtype=self.dtype,
data=Is[i])
recorders = []
for i in record:
recorders.append((i, None))
with open('record_parameters.pickle', 'wb') as f:
pickle.dump(recorders, f, protocol=pickle.HIGHEST_PROTOCOL)
if graph_filename is not None:
nx.write_gexf(self.G, graph_filename)
if execute_in_same_thread:
from neurokernel.core_gpu import Manager
from neurokernel.LPU.LPU import LPU
# import neurokernel.mpi_relaunch
from neurokernel.LPU.InputProcessors.FileInputProcessor import \
FileInputProcessor
from neurokernel.LPU.OutputProcessors.FileOutputProcessor import \
FileOutputProcessor
input_processor = FileInputProcessor(input_filename)
(comp_dict, conns) = LPU.graph_to_dicts(self.G)
output_processor = FileOutputProcessor(
[(i, None) for i in list(record)],
output_filename,
sample_interval=sample_interval)
self.manager = Manager()
self.manager.add(
LPU,
self.experiment_name,
self.config.dt,
comp_dict,
conns,
device=self.config.device,
input_processors=[input_processor],
output_processors=[output_processor],
debug=False,
extra_comps=extra_comps if extra_comps is not None else [])
def setUp(self):
self.man = Manager()
class test_core_gpu(TestCase):
def setUp(self):
self.man = Manager()
def test_transmit_spikes_one_to_one(self):
m1_sel_in_gpot = Selector('')
m1_sel_out_gpot = Selector('')
m1_sel_in_spike = Selector('')
m1_sel_out_spike = Selector('/m1/out/spike[0:4]')
m1_sel, m1_sel_in, m1_sel_out, m1_sel_gpot, m1_sel_spike = \
make_sels(m1_sel_in_gpot, m1_sel_out_gpot, m1_sel_in_spike, m1_sel_out_spike)
N1_gpot = SelectorMethods.count_ports(m1_sel_gpot)
N1_spike = SelectorMethods.count_ports(m1_sel_spike)
m2_sel_in_gpot = Selector('')
m2_sel_out_gpot = Selector('')
m2_sel_in_spike = Selector('/m2/in/spike[0:4]')
m2_sel_out_spike = Selector('')
m2_sel, m2_sel_in, m2_sel_out, m2_sel_gpot, m2_sel_spike = \
make_sels(m2_sel_in_gpot, m2_sel_out_gpot, m2_sel_in_spike, m2_sel_out_spike)
N2_gpot = SelectorMethods.count_ports(m2_sel_gpot)
N2_spike = SelectorMethods.count_ports(m2_sel_spike)
m1_id = 'm1'
self.man.add(MyModule1,
m1_id,
m1_sel,
m1_sel_in,
m1_sel_out,
m1_sel_gpot,
m1_sel_spike,
np.zeros(N1_gpot, dtype=np.double),
np.zeros(N1_spike, dtype=int),
device=0,
debug=debug,
out_spike_data=[0, 0, 1, 1])
f, out_file_name = tempfile.mkstemp()
os.close(f)
m2_id = 'm2'
self.man.add(MyModule2,
m2_id,
m2_sel,
m2_sel_in,
m2_sel_out,
m2_sel_gpot,
m2_sel_spike,
np.zeros(N2_gpot, dtype=np.double),
np.zeros(N2_spike, dtype=int),
device=1,
debug=debug,
out_file_name=out_file_name)
pat12 = Pattern(m1_sel, m2_sel)
pat12.interface[m1_sel_out_gpot] = [0, 'in', 'gpot']
pat12.interface[m1_sel_in_gpot] = [0, 'out', 'gpot']
pat12.interface[m1_sel_out_spike] = [0, 'in', 'spike']
pat12.interface[m1_sel_in_spike] = [0, 'out', 'spike']
pat12.interface[m2_sel_in_gpot] = [1, 'out', 'gpot']
pat12.interface[m2_sel_out_gpot] = [1, 'in', 'gpot']
pat12.interface[m2_sel_in_spike] = [1, 'out', 'spike']
pat12.interface[m2_sel_out_spike] = [1, 'in', 'spike']
pat12['/m1/out/spike[0]', '/m2/in/spike[0]'] = 1
pat12['/m1/out/spike[1]', '/m2/in/spike[1]'] = 1
pat12['/m1/out/spike[2]', '/m2/in/spike[2]'] = 1
pat12['/m1/out/spike[3]', '/m2/in/spike[3]'] = 1
self.man.connect(m1_id, m2_id, pat12, 0, 1)
# Run emulation for 2 steps:
self.man.spawn()
self.man.start(2)
self.man.wait()
# Get output of m2:
with open(out_file_name, 'r') as f:
output = pickle.load(f)
os.remove(out_file_name)
self.assertSequenceEqual(list(output), [0, 0, 1, 1])
def test_transmit_spikes_one_to_many(self):
m1_sel_in_gpot = Selector('')
m1_sel_out_gpot = Selector('')
m1_sel_in_spike = Selector('')
m1_sel_out_spike = Selector('/m1/out/spike[0:4]')
m1_sel, m1_sel_in, m1_sel_out, m1_sel_gpot, m1_sel_spike = \
make_sels(m1_sel_in_gpot, m1_sel_out_gpot, m1_sel_in_spike, m1_sel_out_spike)
N1_gpot = SelectorMethods.count_ports(m1_sel_gpot)
N1_spike = SelectorMethods.count_ports(m1_sel_spike)
m2_sel_in_gpot = Selector('')
m2_sel_out_gpot = Selector('')
m2_sel_in_spike = Selector('/m2/in/spike[0:4]')
m2_sel_out_spike = Selector('')
m2_sel, m2_sel_in, m2_sel_out, m2_sel_gpot, m2_sel_spike = \
make_sels(m2_sel_in_gpot, m2_sel_out_gpot, m2_sel_in_spike, m2_sel_out_spike)
N2_gpot = SelectorMethods.count_ports(m2_sel_gpot)
N2_spike = SelectorMethods.count_ports(m2_sel_spike)
m1_id = 'm1'
self.man.add(MyModule1,
m1_id,
m1_sel,
m1_sel_in,
m1_sel_out,
m1_sel_gpot,
m1_sel_spike,
np.zeros(N1_gpot, dtype=np.double),
np.zeros(N1_spike, dtype=int),
device=0,
debug=debug,
out_spike_data=[1, 0, 0, 0])
f, out_file_name = tempfile.mkstemp()
os.close(f)
m2_id = 'm2'
self.man.add(MyModule2,
m2_id,
m2_sel,
m2_sel_in,
m2_sel_out,
m2_sel_gpot,
m2_sel_spike,
np.zeros(N2_gpot, dtype=np.double),
np.zeros(N2_spike, dtype=int),
device=1,
debug=debug,
out_file_name=out_file_name)
pat12 = Pattern(m1_sel, m2_sel)
pat12.interface[m1_sel_out_gpot] = [0, 'in', 'gpot']
pat12.interface[m1_sel_in_gpot] = [0, 'out', 'gpot']
pat12.interface[m1_sel_out_spike] = [0, 'in', 'spike']
pat12.interface[m1_sel_in_spike] = [0, 'out', 'spike']
pat12.interface[m2_sel_in_gpot] = [1, 'out', 'gpot']
pat12.interface[m2_sel_out_gpot] = [1, 'in', 'gpot']
pat12.interface[m2_sel_in_spike] = [1, 'out', 'spike']
pat12.interface[m2_sel_out_spike] = [1, 'in', 'spike']
pat12['/m1/out/spike[0]', '/m2/in/spike[0]'] = 1
pat12['/m1/out/spike[0]', '/m2/in/spike[1]'] = 1
pat12['/m1/out/spike[0]', '/m2/in/spike[2]'] = 1
pat12['/m1/out/spike[0]', '/m2/in/spike[3]'] = 1
self.man.connect(m1_id, m2_id, pat12, 0, 1)
# Run emulation for 2 steps:
self.man.spawn()
self.man.start(2)
self.man.wait()
# Get output of m2:
with open(out_file_name, 'r') as f:
output = pickle.load(f)
os.remove(out_file_name)
self.assertSequenceEqual(list(output), [1, 1, 1, 1])
