Spinncer update

.gitignore

@@ -1,8 +1,7 @@
-application_generated_data_files/*
-reports/*
+*application_generated_data_files/
+*reports/
 __pycache__/
 .idea/
-spynnaker.cfg
 *.egg-info
 *.npz
-*.out
+*.out

requirements.txt

@@ -1,5 +1,9 @@
-numpy==1.15.4
-scipy==1.3.0
+numpy > 1.13, < 1.20; python_version == '3.6'
+numpy > 1.13, < 1.21; python_version == '3.7'
+numpy; python_version >= '3.8'
+scipy >= 0.16.0, < 1.6; python_version == '3.6'
+scipy >= 0.16.0, < 1.8; python_version == '3.7'
+scipy >= 0.16.0; python_version >= '3.8'
 brian2
 neo==0.7.1
 pynn

setup.py

@@ -2,8 +2,12 @@
 
 
 install_requires = [
-    "numpy==1.15.4",
-    "scipy==1.3.0",
+    "numpy > 1.13, < 1.20; python_version == '3.6'",
+    "numpy > 1.13, < 1.21; python_version == '3.7'",
+    "numpy; python_version >= '3.8'",
+    "scipy >= 0.16.0, < 1.6; python_version == '3.6'",
+    "scipy >= 0.16.0, < 1.8; python_version == '3.7'",
+    "scipy >= 0.16.0; python_version >= '3.8'",
     "brian2",
     "neo==0.7.1",
     "pynn==0.9.5",
@@ -38,7 +42,7 @@
 
         "Programming Language :: Python :: 3"
         "Programming Language :: Python :: 3.7"
-        
+
         "Topic :: Scientific/Engineering",
     ]
 )

spinncer/cerebellum.py

@@ -16,6 +16,9 @@
 import numpy as np
 import h5py
 
+from spynnaker.pyNN.extra_algorithms.splitter_components import (
+    SplitterAbstractPopulationVertexNeuronsSynapses)
+
 from spinncer.analysis_common import get_plot_order
 from spinncer.circuit import Circuit
 from spinncer.utilities.constants import *
@@ -45,6 +48,9 @@ def __init__(self, sim, connectivity, stimulus_information, reporting=True,
                  neuron_model="IF_cond_exp", force_number_of_neurons=None,
                  input_spikes=None,
                  rb_left_shifts=None,
+                 use_split_model=False,
+                 n_synapse_cores=1,
+                 n_delay_slots=64,
                  no_loops=3,
                  round_input_spike_times=None,
                  id_remap=None,
@@ -78,6 +84,11 @@ def __init__(self, sim, connectivity, stimulus_information, reporting=True,
         self.id_remap = id_remap
         self.r_mem = r_mem
 
+        # Values for splitters
+        self.use_split_model = use_split_model
+        self.n_synapse_cores = n_synapse_cores
+        self.n_delay_slots = n_delay_slots
+
         self.expected_max_spikes = expected_max_spikes
         self.implicit_shift = implicit_shift
         self.ensure_weight_is_representable = ensure_weight_is_representable
@@ -475,6 +486,18 @@ def build_populations(self, positions):
                 elif cell_model == "if_cond_alpha":
                     cell_model = self.sim.IF_cond_alpha
 
+            # Add splitters everywhere except for SSP populations
+            if self.use_split_model:
+                if cell_name in ["granule", "golgi", "stellate", "basket",
+                                 "purkinjie", "dcn"]:
+                    print("Cell {} model {} using split synapse neuron model "
+                          "with {} synapse cores and {} delay slots".format(
+                              cell_name, cell_model, self.n_synapse_cores,
+                              self.n_delay_slots))
+                    additional_params["splitter"] = \
+                        SplitterAbstractPopulationVertexNeuronsSynapses(
+                            self.n_synapse_cores, self.n_delay_slots, False)
+
             # Adding the population to the network
             try:
                 self.populations[cell_name] = self.sim.Population(

spinncer/cerebellum_analysis.py

@@ -417,6 +417,8 @@ def spike_analysis(results_file, fig_folder,
     try:
         for pop in plot_order:
             curr_spikes = neo_all_spikes[pop].segments[0].spiketrains
+            print("pop, curr_spikes: ", pop, curr_spikes)
+            print("elephant, timestep, simtime", elephant_timestep, elephant_simtime)
             curr_inst_rates = \
                 elephant.statistics.instantaneous_rate(
                     curr_spikes,
@@ -437,6 +439,9 @@ def spike_analysis(results_file, fig_folder,
     except:
         traceback.print_exc()
 
+    print("elephant_instantaneous_rates: ", elephant_instantaneous_rates)
+    print("all_neurons: ", all_neurons)
+
     stim_period_start = {}
     stim_period_end = {}
     per_pop_stim_durations = {k: [] for k in plot_order}
@@ -465,6 +470,7 @@ def spike_analysis(results_file, fig_folder,
         # before, during and after stimulation
         _filtered_spike_rates = np.zeros(stimulus_periods)
         _spike_times = spikes[:, 1]
+        _spike_nids = spikes[:, 0]
         # Initialise per_neuron_firing
         per_neuron_firing[pop] = np.ones((all_neurons[pop],
                                           stimulus_periods)) * -10
@@ -498,7 +504,7 @@ def spike_analysis(results_file, fig_folder,
                 np.count_nonzero(_filtered_spike_times) / \
                 (current_period_duration * ms)
             for nid in range(all_neurons[pop]):
-                _spikes_for_nid = spikes[spikes[:, 0] == nid][:, 1]
+                _spikes_for_nid = spikes[_spike_nids == nid][:, 1]
                 _no_spike_for_nid = np.count_nonzero(np.logical_and(
                     _spikes_for_nid >= time_filter_pre,
                     _spikes_for_nid < time_filter_post))
@@ -712,8 +718,10 @@ def spike_analysis(results_file, fig_folder,
     print("Connection Name ")
     print("{:27} | {:10} | ".format("Connection Name", "Def. W"),
           "{:20}".format("SpiNN. W"))
-    sorted_keys = list(final_connectivity.keys())
-    sorted_keys.sort()
+    sorted_keys = []
+    if final_connectivity != []:
+        sorted_keys = list(final_connectivity.keys())
+        sorted_keys.sort()
     for key in sorted_keys:
         conn = conn_dict[key]
         mean = np.abs(np.mean(conn[:, 2]))
@@ -802,8 +810,8 @@ def spike_analysis(results_file, fig_folder,
         # Report statistics here
         for key, v in all_voltages.items():
             nid, tstep = np.unravel_index(np.argmax(v, axis=None), v.shape)
-            print("{:20}-> neuron {:>8d} received {:>6d}".format(
-                key, int(nid), int(np.max(v))),
+            print("{:20}-> neuron {:>8d} received {:4.2f}".format(
+                key, int(nid), np.max(v)),
                 "nA in timestep #{:8d}".format(int(tstep)))
             # THIS IS BROKEN! it will be removed soon
             # # Also treat voltage as if it's a piggybacked value packaging

spinncer/cerebellum_experiment.py

@@ -12,11 +12,7 @@
 # import simulator
 spinnaker_sim = False
 if str.lower(args.simulator) in ["spinnaker", "spynnaker"]:
-    try:
-        # this might be deprecated soon
-        import spynnaker8 as sim
-    except ImportError:
-        import pyNN.spynnaker as sim
+    import pyNN.spiNNaker as sim
     spinnaker_sim = True
 elif str.lower(args.simulator) in ["nest"]:
     import pyNN.nest as sim
@@ -32,6 +28,7 @@
 import pylab as plt
 import os
 import traceback
+from spinncer.provenance_analysis import save_provenance_to_file_from_database
 
 # Record SCRIPT start time (wall clock)
 start_time = plt.datetime.datetime.now()
@@ -137,6 +134,8 @@
 expected_max_spikes = EXPECTED_MAX_SPIKES_200
 canonical_rbls = RMEM_RBLS if args.r_mem else VANILLA_RBLS
 print("Canonical ring buffer left shifts:", canonical_rbls)
+use_split_model = True
+print("use_split_model is ", use_split_model)
 
 # Instantiate a Cerebellum
 cerebellum_circuit = Cerebellum(
@@ -149,6 +148,9 @@
     neuron_model=args.neuron_model,
     input_spikes=input_spikes,
     rb_left_shifts=canonical_rbls,
+    use_split_model=use_split_model,
+    n_synapse_cores=1,
+    n_delay_slots=64,
     no_loops=args.loops_grc,
     round_input_spike_times=round_spike_times,
     id_remap=args.id_remap,
@@ -234,6 +236,9 @@
 recorded_spikes = {}
 other_recordings = {}
 
+# Store simulator name and run
+sim_name = sim.name
+
 # Record simulation start time (wall clock)
 sim_start_time = plt.datetime.datetime.now()
 current_error = None
@@ -311,6 +316,11 @@
 # Appropriately end the simulation
 sim.end()
 
+# Get the provenance from the simulator's database
+structured_provenance_filename = "structured_provenance.csv"
+save_provenance_to_file_from_database(
+    structured_provenance_filename, sim_name)
+
 # Analysis time!
 spike_analysis(results_file=results_file, fig_folder=args.figures_dir,
                worst_case=args.worst_case_spikes, delay_sensitive=True)