rowleya
commented
2 years ago I am struggling to replicate this; can you provide your whole script?
Open Helloworld-Le opened 2 years ago
rowleya
commented
2 years ago I am struggling to replicate this; can you provide your whole script?
Helloworld-Le
commented
2 years ago here is the code:
`# Import the simulator
import pyNN.spiNNaker as p
import numpy
from pyNN.space import Grid2D
# Import the extra classes
from spynnaker.pyNN.external_devices_models import (
AbstractEthernetTranslator, AbstractMulticastControllableDevice)
from spynnaker.pyNN.external_devices_models \
.abstract_multicast_controllable_device import SendType
from data_specification.enums import DataType
class TestTranslator(AbstractEthernetTranslator):
def translate_control_packet(self, packet):
print("Received key={}, voltage={}".format(
packet.key, packet.payload / DataType.S1615.max))
class TestDevice(AbstractMulticastControllableDevice):
@property
def device_control_partition_id(self):
# This should be unique to the device, but is otherwise unimportant
return "Test"
@property
def device_control_key(self):
# This should be unique to the device
return 0
@property
def device_control_uses_payload(self):
# This returns True to receive the voltage,
# or False if only the key is desired
return True
@property
def device_control_min_value(self):
# Return the minimum value accepted by the device. If the membrane
# voltage is below this value, this value will be used.
return 0
@property
def device_control_max_value(self):
# Return the maximum value accepted by the device. If the membrane
# voltage is above this value, this value will be used.
return 100
@property
def device_control_timesteps_between_sending(self):
# The number of timesteps between sending values. Controls the
# update rate of the value.
return 10
@property
def device_control_send_type(self):
# The type of the value - one of the SendType values
return SendType.SEND_TYPE_ACCUM
@property
def device_control_scaling_factor(self):
# The amount to multiply the voltage by before transmission
return 1.0
# Setup the simulation
p.setup(1.0)
# Run time if send_fake_spikes is False
run_time = 60000
# Constants
P_SHIFT = 15
Y_SHIFT = 0
X_SHIFT = 16
WIDTH = 346
HEIGHT = 260
# Creates 512 neurons per core
SUB_WIDTH = 16
SUB_HEIGHT = 16
# Weight of connections between "layers"
WEIGHT = 1
# Set the number of neurons per core to a rectangle
# (creates 512 neurons per core)
p.set_number_of_neurons_per_core(p.IF_curr_exp, (SUB_WIDTH, SUB_HEIGHT))
#capture_conn = p.ConvolutionConnector([[WEIGHT]])
# These are our external retina devices connected to SPIF devices
devices = list()
captures = list()
dev = p.Population(None, p.external_devices.SPIFRetinaDevice(
pipe=0, width=WIDTH, height=HEIGHT, sub_width=SUB_WIDTH,
sub_height=SUB_HEIGHT, input_x_shift=X_SHIFT, input_y_shift=Y_SHIFT,
))
# Make a kernel and convolution connector
k_shape = numpy.array([5, 5], dtype='int32')
k_size = numpy.prod(k_shape)
kernel = (numpy.arange(k_size) - (k_size / 2)).reshape(k_shape) * 0.1
conn = p.ConvolutionConnector(kernel)
# Start with an input shape, and deduce the output shape
in_shape = (WIDTH, HEIGHT)
out_shape = conn.get_post_shape(in_shape)
n_input = numpy.prod(in_shape)
n_output = numpy.prod(out_shape)
# Make a 2D target Population and record it
capture = p.Population(
(out_shape[0] * out_shape[1]), p.IF_curr_exp(), label="out",
structure=Grid2D(out_shape[0] / out_shape[1]))
p.Projection(dev, capture, conn, p.Convolution())
# Create the model that will generate the voltage
pop = p.external_devices.EthernetControlPopulation(
n_neurons=1,
model=p.external_devices.ExternalDeviceLifControl(
devices=[TestDevice()],
create_edges=False,
translator=TestTranslator()),
label="test")
# Connect the stimulation to the population
p.Projection(capture, pop, p.OneToOneConnector(), p.StaticSynapse(weight=0.1))
# Run the simulation then stop
p.run(run_time)
p.end()`I believe that this is now fixed on the extdev_fpgas branch. The changes are in SpiNNFrontEndCommon - please make sure that you pull extdev_fpgas on that repository before trying again.
Helloworld-Le
commented
2 years ago **Hi, Now there is no error as showing before, but I don't see the voltage being printed here.
To update, now my whole script is :**
`
# Import the simulator
import pyNN.spiNNaker as p
import numpy
from pyNN.space import Grid2D
import pyNN.utility.plotting as plot
from datetime import datetime
from spynnaker.pyNN.external_devices_models import AbstractMulticastControllableDevice
from spynnaker.pyNN.external_devices_models.abstract_multicast_controllable_device import SendType
from spynnaker.pyNN.external_devices_models import AbstractEthernetTranslator
#outputpath = ''
figoutputpath = '/Users/Nono/PycharmProjects/SpiNNaker_test/output_data/plots/'
recordoutputpath = '/Users/Nono/PycharmProjects/SpiNNaker_test/output_data/record/'
class Translator(AbstractEthernetTranslator):
def translate_control_packet(self, packet):
print("Received Key {}: Voltage {}".format(packet.key, packet.payload))
class TestDevice(AbstractMulticastControllableDevice):
@property
def device_control_partition_id(self):
return "Test"
@property
def device_control_key(self):
return 0
@property
def device_control_uses_payload(self):
return True
@property
def device_control_min_value(self):
return -100
@property
def device_control_max_value(self):
return 100
@property
def device_control_timesteps_between_sending(self):
return 10
@property
def device_control_send_type(self):
# The type of the value - one of the SendType values
return SendType.SEND_TYPE_UINT
@property
def device_control_scaling_factor(self):
# The amount to multiply the voltage by before transmission
return 1.0
p.setup(3.0)
run_time= 10000
# Constants
P_SHIFT = 15
Y_SHIFT = 0
X_SHIFT = 16
WIDTH = 346
HEIGHT = 260
# Creates 512 neurons per core
SUB_WIDTH = 16
SUB_HEIGHT = 16
# Weight of connections between "layers"
WEIGHT = 1
# Set the number of neurons per core to a rectangle
# (creates 512 neurons per core)
p.set_number_of_neurons_per_core(p.IF_curr_exp, (SUB_WIDTH, SUB_HEIGHT))
#capture_conn = p.ConvolutionConnector([[WEIGHT]])
# These are our external retina devices connected to SPIF devices
devices = list()
captures = list()
dev = p.Population(None, p.external_devices.SPIFRetinaDevice(
pipe=0, width=WIDTH, height=HEIGHT, sub_width=SUB_WIDTH,
sub_height=SUB_HEIGHT, input_x_shift=X_SHIFT, input_y_shift=Y_SHIFT,
))
# Make a kernel and convolution connector
k_shape = numpy.array([5, 5], dtype='int32')
k_size = numpy.prod(k_shape)
kernel = (numpy.arange(k_size) - (k_size / 2)).reshape(k_shape) * 0.1
conn = p.ConvolutionConnector(kernel)
#print(k_size,kernel,conn)
# Start with an input shape, and deduce the output shape
in_shape = (WIDTH, HEIGHT)
out_shape = conn.get_post_shape(in_shape)
#print(out_shape)
n_input = numpy.prod(in_shape)
n_output = numpy.prod(out_shape)
# Make a 2D target Population and record it
capture = p.Population(
(out_shape[0] * out_shape[1]), p.IF_curr_exp(), label="out",
structure=Grid2D(out_shape[0] / out_shape[1]))
p.Projection(dev, capture, conn, p.Convolution())
#source = p.Population(100, p.SpikeSourcePoisson(rate=10), label="source")
pop = p.external_devices.EthernetControlPopulation(
n_neurons=1,
model=p.external_devices.ExternalDeviceLifControl(
devices=[TestDevice()],
create_edges=False,
translator=Translator()))
p.Projection(capture, pop, p.AllToAllConnector(), p.StaticSynapse(weight=0.1, delay=1))
capture.record("spikes")
# Run the simulation then stop
p.run(run_time)
record_spikes = capture.get_data(variables=["spikes"]).segments[0].spiketrains
p.end()
plot.Figure(
# plot spikes (or in this case spike)
plot.Panel(record_spikes, xticks =True, yticks=True, markersize=1, xlim=(0,run_time)),
#plot.Panel(pn_spikes, xticks =True, yticks=True, markersize=1, xlim=(0,self.sim_t)),
).save(figoutputpath+datetime.now().strftime('%Y_%m_%d_%H_%M_%S')+"record.png")
`not too much changed, but just to add a spike record of the capture layer. From this record I can see some spikes, which is from the DVS camera. Then the whole running report is like:
2022-05-13 13:24:26 INFO: Read configs files: /Users/Nono/PycharmProjects/CCNW2022/SpiNNUtils/spinn_utilities/spinn_utilities.cfg, /Users/Nono/PycharmProjects/CCNW2022/SpiNNMachine/spinn_machine/spinn_machine.cfg, /Users/Nono/PycharmProjects/CCNW2022/PACMAN/pacman/pacman.cfg, /Users/Nono/PycharmProjects/CCNW2022/SpiNNMan/spinnman/spinnman.cfg, /Users/Nono/PycharmProjects/CCNW2022/DataSpecification/data_specification/data_specification.cfg, /Users/Nono/PycharmProjects/CCNW2022/SpiNNFrontEndCommon/spinn_front_end_common/interface/spinnaker.cfg, /Users/Nono/PycharmProjects/CCNW2022/sPyNNaker/spynnaker/pyNN/spynnaker.cfg, /Users/Nono/.spynnaker.cfg
2022-05-13 13:24:26 INFO: Will search these locations for binaries: /Users/Nono/PycharmProjects/CCNW2022/SpiNNFrontEndCommon/spinn_front_end_common/common_model_binaries : /Users/Nono/PycharmProjects/CCNW2022/sPyNNaker/spynnaker/pyNN/model_binaries
2022-05-13 13:24:26 WARNING: /Users/Nono/PycharmProjects/SpiNNaker_test/reports has 95 old reports that have not been closed
2022-05-13 13:24:26 INFO: Setting time scale factor to 1.0.
2022-05-13 13:24:26 INFO: Setting machine time step to 3000 micro-seconds.
2022-05-13 13:24:26 WARNING: Size of the population out rounded from 87552 to 87552. Please use int values for size
2022-05-13 13:24:26 WARNING: Formal PyNN specifies that devices should be set using initial_values not cell_params
2022-05-13 13:24:26 WARNING: Formal PyNN specifies that create_edges should be set using initial_values not cell_params
2022-05-13 13:24:26 WARNING: Formal PyNN specifies that translator should be set using initial_values not cell_params
2022-05-13 13:24:26 INFO: 0.0.0.0:63012 Waiting for message to indicate that the database is ready
2022-05-13 13:24:26 INFO: Receive callback <bound method EthernetControlConnection._translate of UDPConnection(local=0.0.0.0:63012, remote=None:None)> registered to label ExternalDeviceLifControlVertex_2
2022-05-13 13:24:26 INFO: Starting execution process
2022-05-13 13:24:26 WARNING: Time of 10000.0ms is not a multiple of the machine time step of 3.0ms and has therefore been rounded up to 10002.0ms
2022-05-13 13:24:26 INFO: Simulating for 3334 3.0ms timesteps using a hardware timestep of 3000.0us
2022-05-13 13:24:26 INFO: Creating transceiver for 172.16.223.43
2022-05-13 13:24:26 INFO: Working out if machine is booted
2022-05-13 13:24:26 INFO: Found board with version [Version: SC&MP 3.4.2 at SpiNNaker:0:0:0 (built Fri Feb 11 08:07:06 2022)]
2022-05-13 13:24:26 INFO: Machine communication successful
2022-05-13 13:24:26 INFO: 172.16.223.43
2022-05-13 13:24:26 INFO: Detected a machine on IP address 172.16.223.43 which has 858 cores and 120.0 links
2022-05-13 13:24:26 INFO: Machine generator took 0:00:00.059036
2022-05-13 13:24:26 INFO: SpYNNakerNeuronGraphNetworkSpecificationReport skipped as cfg Reports:write_network_graph is False
2022-05-13 13:24:26 INFO: Network Specification report took 0:00:00.000405
2022-05-13 13:24:26 INFO: Splitter reset took 0:00:00.000013
Adding Splitter selectors where appropriate
|0% 50% 100%|
============================================================
2022-05-13 13:24:26 INFO: Spynnaker splitter selector took 0:00:00.031596
Adding delay extensions as required
|0% 50% 100%|
============================================================
2022-05-13 13:24:26 INFO: DelaySupportAdder took 0:00:00.000256
2022-05-13 13:24:26 INFO: Preallocate for live packet gatherer took 0:00:00.000281
2022-05-13 13:24:26 INFO: Preallocate for chip power monitor skipped as cfg Reports:write_energy_report is False
2022-05-13 13:24:26 INFO: Preallocate for extra monitor support took 0:00:00.000053
Partitioning graph vertices
|0% 50% 100%|
============================================================
Partitioning graph edges
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: SpynnakerSplitterPartitioner took 0:00:27.490579
2022-05-13 13:24:54 INFO: Json machine skipped as cfg Reports:write_json_machine is False
Allocating virtual identifiers
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Chip ID allocator took 0:00:00.036095
Adding Live Packet Gatherers to Graph
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Insert live packet gatherers to graphs took 0:00:00.000828
Writing the board chip report
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Board chip report took 0:00:00.000671
2022-05-13 13:24:54 INFO: Insert chip power monitors skipped as cfg Reports:write_energy_report is False
2022-05-13 13:24:54 INFO: Insert extra monitor vertices took 0:00:00.000022
Inserting extra monitors into graphs
|0% 50% 100%|
============================================================
Generating partitioner report
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Partitioner report took 0:00:00.013655
Getting number of keys required by each edge using application graph
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Edge to n keys mapper took 0:00:00.021544
2022-05-13 13:24:54 INFO: Local TDMA builder took 0:00:00.003095
2022-05-13 13:24:54 INFO: Json partition n keys map skipped as cfg Reports:write_json_partition_n_keys_map is False
Placing graph vertices via spreading over an entire machine
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Spreader placer took 0:00:00.177729
Adding edges to the machine graph between the vertices to which live output has been requested and its local Live Packet Gatherer
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Insert edges to live packet gatherers took 0:00:00.000413
Inserting edges between vertices which require FR speed up functionality.
|0% 50% 100%|
===========================================================2022-05-13 13:24:54 INFO: Insert Edges To Extra Monitor took 0:00:00.009051
Generating routing tables for data in system processes
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: System multicast routing generator took 0:00:00.001937
Generating fixed router routes
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Fixed route router took 0:00:00.000714
Generating placement report
|0% 50% 100%|
============================================================
Generating placement by core report
|0% 50% 100%|
============================================================
2022-05-13 13:24:54 INFO: Placements wth application graph report took 0:00:00.023805
2022-05-13 13:24:54 INFO: Placements wthout machine graaph skipped as cfg Reports:write_machine_graph_placer_report is False
2022-05-13 13:24:54 INFO: Json placements skipped as cfg Reports:write_json_placements is False
Routing
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Ner route traffic aware took 0:00:00.434253
Discovering tags
|0% 50% 100%|
============================================================
Allocating tags
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Basic tag allocator took 0:00:00.013646
Reporting Tags
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Tag allocator report took 0:00:00.000408
Getting constraints for machine graph
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Process partition constraints took 0:00:00.152017
Calculating zones
|0% 50% 100%|
============================================================
Allocating routing keys
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Zoned routing info allocator took 0:00:00.294108
Generating Routing info report
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Router info report took 0:00:00.031645
Generating routing tables
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Basic routing table generator took 0:00:00.217119
2022-05-13 13:24:55 INFO: Router report skipped as cfg Reports:write_router_reports is False
2022-05-13 13:24:55 INFO: Router summary report skipped as cfg Reports:write_router_summary_report is False
2022-05-13 13:24:55 INFO: Json routing tables skipped as cfg Reports:write_json_routing_tables is False
Finding executable start types
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Locate executable start type took 0:00:00.009703
Initialising buffers
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: Buffer manager creator took 0:00:00.016192
Allocating SDRAM for SDRAM outgoing egde partitions
|0% 50% 100%|
============================================================
2022-05-13 13:24:55 INFO: SDRAM outgoing partition allocator took 0:00:00.006147
Generating data specifications
|0% 50% 100%|
============================================================
2022-05-13 13:25:18 INFO: Spynnaker data specification writer took 0:00:22.785341
Preparing Routing Tables
|0% 50% 100%|
============================================================
2022-05-13 13:25:19 INFO: Routing setup took 0:00:00.492857
Finding binaries
|0% 50% 100%|
============================================================
2022-05-13 13:25:19 INFO: Graph binary gatherer took 0:00:00.022456
Generating Router table report
|0% 50% 100%|
============================================================
2022-05-13 13:25:19 INFO: Uncompressed routing table report took 0:00:00.238197
2022-05-13 13:25:19 INFO: Ranged Compressor skipped as Tables already small enough
2022-05-13 13:25:19 INFO: Pair on chip router compression skipped as Tables already small enough
loading fixed routes
|0% 50% 100%|
============================================================
2022-05-13 13:25:19 INFO: Load fixed routes took 0:00:00.151860
Executing data specifications and loading data for system vertices
|0% 50% 100%|
============================================================
2022-05-13 13:25:20 INFO: Execute system data specification took 0:00:00.745937
Loading system executables onto the machine
|0% 50% 100%|
============================================================
2022-05-13 13:25:26 INFO: Load executable system Images took 0:00:06.014236
Clearing tags
|0% 50% 100%|
============================================================
Loading Tags
|0% 50% 100%|
============================================================
2022-05-13 13:25:26 INFO: Tags Loader took 0:00:00.029973
Executing data specifications and loading data for application vertices
|0% 50% 100%|
============================================================
2022-05-13 13:26:43 INFO: Host data specification took 0:01:17.603319
Preparing to Expand Synapses
|0% 50% 100%|
============================================================
Expanding Synapses
|0% 50% 100%|
============================================================
2022-05-13 13:26:48 INFO: Synapse expander took 0:00:04.579483
Running bitfield generation on chip
|0% 50% 100%|
============================================================
2022-05-13 13:26:52 INFO: Execute on chip bitfield generator took 0:00:03.908035
Finalising Retrieved Connections
|0% 50% 100%|
============================================================
2022-05-13 13:26:52 INFO: Finish connection holders took 0:00:00.000399
Loading routing data onto the machine
|0% 50% 100%|
============================================================
2022-05-13 13:26:53 INFO: Routing table loader took 0:00:01.041134
2022-05-13 13:26:53 INFO: Bitfield compressor report skipped as cfg Reports:write_bit_field_compressor_report is False
2022-05-13 13:26:53 INFO: Tags from machine report took 0:00:00.010414
2022-05-13 13:26:53 INFO: Memory report skipped as cfg Reports:write_memory_map_report is False
2022-05-13 13:26:53 INFO: Memory report skipped as cfg Reports:write_memory_map_report is False
Generating compressed router table report
|0% 50% 100%|
============================================================
Generating comparison of router table report
|0% 50% 100%|
============================================================
Generating Routing summary report
|0% 50% 100%|
============================================================
Reading Routing Tables from Machine
|0% 50% 100%|
============================================================
2022-05-13 13:26:54 INFO: Compressor report took 0:00:00.600828
Writing fixed route report
|0% 50% 100%|
============================================================
2022-05-13 13:26:54 INFO: Fixed route report took 0:00:00.078317
Loading executables onto the machine
|0% 50% 100%|
============================================================
2022-05-13 13:27:07 INFO: Load executable app images took 0:00:13.607111
2022-05-13 13:27:07 INFO: Running for 1 steps for a total of 10000.0ms
2022-05-13 13:27:07 INFO: Run 1 of 1
Generating SDRAM usage report
|0% 50% 100%|
============================================================
2022-05-13 13:27:07 INFO: Sdram usage per chip report took 0:00:00.080599
2022-05-13 13:27:07 INFO: creating live event connection database in /Users/Nono/PycharmProjects/SpiNNaker_test/reports/2022-05-13-13-24-26-582724/run_1/input_output_database.sqlite3
Creating graph description database
|0% 50% 100%|
============================================================
2022-05-13 13:27:09 INFO: Create database interface took 0:00:01.864240
2022-05-13 13:27:09 INFO: ** Notifying external sources that the database is ready for reading **
2022-05-13 13:27:09 INFO: 0.0.0.0:63012 Reading database
2022-05-13 13:27:09 INFO: database is at /Users/Nono/PycharmProjects/SpiNNaker_test/reports/2022-05-13-13-24-26-582724/run_1/input_output_database.sqlite3
2022-05-13 13:27:09 INFO: Create notification protocol took 0:00:00.002243
Waiting for cores to be either in PAUSED or READY state
|0% 50% 100%|
2022-05-13 13:27:09 INFO: Listening for traffic from ExternalDeviceLifControlVertex_2 on board 172.16.223.43 on 0.0.0.0:55159
2022-05-13 13:27:09 INFO: Notifying the toolchain that the database has been read
2022-05-13 13:27:09 INFO: Waiting for message to indicate that the simulation has started or resumed
2022-05-13 13:27:09 INFO: ** Confirmation from 127.0.0.1:63012 received, continuing **
============================================================
Updating run time
|0% 50% 100%|
============================================================
2022-05-13 13:27:10 INFO: Runtime Update took 0:00:00.307206
2022-05-13 13:27:10 INFO: *** Running simulation... ***
Loading buffers
|0% 50% 100%|
============================================================
2022-05-13 13:27:10 INFO: ** Awaiting for a response from an external source to state its ready for the simulation to start **
2022-05-13 13:27:10 INFO: ** Sending start / resume message to external sources to state the simulation has started or resumed. **
2022-05-13 13:27:10 INFO: ** Awaiting for a response from an external source to state its ready for the simulation to start **
2022-05-13 13:27:10 INFO: Application started; waiting 10.102s for it to stop
2022-05-13 13:27:10 INFO: Waiting for message to indicate that the simulation has stopped or paused
2022-05-13 13:27:20 INFO: ** Sending pause / stop message to external sources to state the simulation has been paused or stopped. **
2022-05-13 13:27:20 INFO: Application runner took 0:00:10.377314
2022-05-13 13:27:20 INFO: Extract IO buff skipped as cfg Reports:extract_iobuf is False
Extracting buffers from the last run
|0% 50% 100%|
============================================================
2022-05-13 13:27:27 INFO: Buffer extractor took 0:00:07.221020
clearing IOBUF from the machine
|0% 50% 100%|
============================================================
2022-05-13 13:27:27 INFO: Clear IO buffer took 0:00:00.206579
Getting provenance data from machine graph
|0% 50% 100%|
============================================================
Getting provenance data from application graph
|0% 50% 100%|
============================================================
2022-05-13 13:27:27 INFO: Graph provenance gatherer took 0:00:00.039218
Getting provenance data
|0% 50% 100%|
=====================================================2022-05-13 13:27:28 WARNING: On Slice Slice(lo_atom=0, hi_atom=255, n_atoms=256, shape=(16, 16), start=(0, 0)) of out on 4,4,16, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=82080, hi_atom=82335, n_atoms=256, shape=(16, 16), start=(0, 240)) of out on 4,4,15, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=87456, hi_atom=87551, n_atoms=96, shape=(6, 16), start=(336, 240)) of out on 4,4,14, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=87200, hi_atom=87455, n_atoms=256, shape=(16, 16), start=(320, 240)) of out on 4,4,13, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=86944, hi_atom=87199, n_atoms=256, shape=(16, 16), start=(304, 240)) of out on 4,4,12, 7 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=86688, hi_atom=86943, n_atoms=256, shape=(16, 16), start=(288, 240)) of out on 4,4,11, 5 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=86432, hi_atom=86687, n_atoms=256, shape=(16, 16), start=(272, 240)) of out on 4,4,10, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
=2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=84384, hi_atom=84639, n_atoms=256, shape=(16, 16), start=(144, 240)) of out on 4,4,2, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=84128, hi_atom=84383, n_atoms=256, shape=(16, 16), start=(128, 240)) of out on 4,4,1, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=82848, hi_atom=83103, n_atoms=256, shape=(16, 16), start=(48, 240)) of out on 5,4,12, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=82592, hi_atom=82847, n_atoms=256, shape=(16, 16), start=(32, 240)) of out on 5,4,11, 5 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=82336, hi_atom=82591, n_atoms=256, shape=(16, 16), start=(16, 240)) of out on 5,4,10, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=81984, hi_atom=82079, n_atoms=96, shape=(6, 16), start=(336, 224)) of out on 5,4,9, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=81728, hi_atom=81983, n_atoms=256, shape=(16, 16), start=(320, 224)) of out on 5,4,8, 4 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=81472, hi_atom=81727, n_atoms=256, shape=(16, 16), start=(304, 224)) of out on 5,4,7, 8 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=81216, hi_atom=81471, n_atoms=256, shape=(16, 16), start=(288, 224)) of out on 5,4,6, 9 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:29 WARNING: On Slice Slice(lo_atom=80960, hi_atom=81215, n_atoms=256, shape=(16, 16), start=(272, 224)) of out on 5,4,5, 2 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:30 WARNING: On Slice Slice(lo_atom=79680, hi_atom=79935, n_atoms=256, shape=(16, 16), start=(192, 224)) of out on 5,5,16, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:30 WARNING: On Slice Slice(lo_atom=78912, hi_atom=79167, n_atoms=256, shape=(16, 16), start=(144, 224)) of out on 5,5,13, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
2022-05-13 13:27:30 WARNING: Additional interesting provenace items in /Users/Nono/PycharmProjects/SpiNNaker_test/reports/2022-05-13-13-24-26-582724/run_1/provenance_data/provenance.sqlite3
======
2022-05-13 13:27:42 INFO: Placements provenance gatherer took 0:00:14.082447
Getting Router Provenance
|0% 50% 100%|
============================================================
2022-05-13 13:27:44 INFO: Router provenance gatherer took 0:00:02.709061
Getting profile data
|0% 50% 100%|
============================================================
2022-05-13 13:27:48 INFO: Profile data gatherer took 0:00:03.324960
2022-05-13 13:27:48 INFO: Energy report skipped as cfg Reports:write_energy_report is False
2022-05-13 13:27:48 INFO: Redundant packet count report took 0:00:00.016680
Getting spikes for out
|0% 50% 100%|
============================================================
These log messages where generated at level WARNING or above
/Users/Nono/PycharmProjects/SpiNNaker_test/reports has 95 old reports that have not been closed
Size of the population out rounded from 87552 to 87552. Please use int values for size
Formal PyNN specifies that devices should be set using initial_values not cell_params
Formal PyNN specifies that create_edges should be set using initial_values not cell_params
Formal PyNN specifies that translator should be set using initial_values not cell_params
Time of 10000.0ms is not a multiple of the machine time step of 3.0ms and has therefore been rounded up to 10002.0ms
On Slice Slice(lo_atom=0, hi_atom=255, n_atoms=256, shape=(16, 16), start=(0, 0)) of out on 4,4,16, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=82080, hi_atom=82335, n_atoms=256, shape=(16, 16), start=(0, 240)) of out on 4,4,15, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=87456, hi_atom=87551, n_atoms=96, shape=(6, 16), start=(336, 240)) of out on 4,4,14, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=87200, hi_atom=87455, n_atoms=256, shape=(16, 16), start=(320, 240)) of out on 4,4,13, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=86944, hi_atom=87199, n_atoms=256, shape=(16, 16), start=(304, 240)) of out on 4,4,12, 7 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=86688, hi_atom=86943, n_atoms=256, shape=(16, 16), start=(288, 240)) of out on 4,4,11, 5 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=86432, hi_atom=86687, n_atoms=256, shape=(16, 16), start=(272, 240)) of out on 4,4,10, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=84384, hi_atom=84639, n_atoms=256, shape=(16, 16), start=(144, 240)) of out on 4,4,2, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=84128, hi_atom=84383, n_atoms=256, shape=(16, 16), start=(128, 240)) of out on 4,4,1, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=82848, hi_atom=83103, n_atoms=256, shape=(16, 16), start=(48, 240)) of out on 5,4,12, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=82592, hi_atom=82847, n_atoms=256, shape=(16, 16), start=(32, 240)) of out on 5,4,11, 5 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=82336, hi_atom=82591, n_atoms=256, shape=(16, 16), start=(16, 240)) of out on 5,4,10, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=81984, hi_atom=82079, n_atoms=96, shape=(6, 16), start=(336, 224)) of out on 5,4,9, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=81728, hi_atom=81983, n_atoms=256, shape=(16, 16), start=(320, 224)) of out on 5,4,8, 4 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=81472, hi_atom=81727, n_atoms=256, shape=(16, 16), start=(304, 224)) of out on 5,4,7, 8 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=81216, hi_atom=81471, n_atoms=256, shape=(16, 16), start=(288, 224)) of out on 5,4,6, 9 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=80960, hi_atom=81215, n_atoms=256, shape=(16, 16), start=(272, 224)) of out on 5,4,5, 2 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=79680, hi_atom=79935, n_atoms=256, shape=(16, 16), start=(192, 224)) of out on 5,5,16, 1 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
On Slice Slice(lo_atom=78912, hi_atom=79167, n_atoms=256, shape=(16, 16), start=(144, 224)) of out on 5,5,13, 3 packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method.
Additional interesting provenace items in /Users/Nono/PycharmProjects/SpiNNaker_test/reports/2022-05-13-13-24-26-582724/run_1/provenance_data/provenance.sqlite3
Process finished with exit code 0Apart form the missing voltage, I also see the 'packets arrived too late to be processed in a given time step. Try increasing the time_scale_factor located within the .spynnaker.cfg file or in the pynn.setup() method' warning as we talked before. In my .cfg file I already added [Simulation] drop_late_spikes = False
best, Le
Helloworld-Le
commented
2 years ago Hi, @rowleya after updating to the newest version, I am getting this new error (with the same code):
The transmission buffer for CommandSender for SPIFRetinaDevice_1 on 0,0,17 was blocked on 56 occasions. This is often a sign that the system is experiencing back pressure from the communication fabric. Please either: 1. spread the load over more cores, 2. reduce your peak transmission load, or 3. adjust your mapping algorithm.
The router on 0, 0 has dropped 124 multicast packets that were transmitted by local cores. This occurs where the router has no entry associated with the multicast key. Try investigating the keys allocated to the vertices and the router table entries for this chip.
The extra monitor on 0, 0 has detected that 1025015 packets were dumped from an outgoing link of this chip's router. This often occurs when external devices are used in the script but not connected to the communication fabric correctly. These packets may have been reinjected multiple times and so this number may be an overestimate.
Process finished with exit code 0
From my recording plot, the capture layer has no spikes.
I am having a DAVIS346 (346*240) and I am also wondering how should I change the 'mapping algorithm' to "collapse" event addresses to a smaller "plane of neurons".
Thanks!
rowleya
commented
2 years ago I think I missed the previous update; that one is indicating that a few packets are still unprocessed at the end of the time step. This is similar to previously with drop_late_spikes=False. The warning is still there to let you know, but those spikes will be processed in the next time step, which is likely not going to make much of a difference in this case. It happens because the spikes likely arrive late in the time step, so there is not enough time to process them. This is because the retina / spif doesn't maintain a synchronized time step with the board.
In the last message, this seems surprising. It may be that the spif board and SpiNNaker boards both need to be reset in that case (remove power from both, wait a few seconds and then plug them in again). Hopefully this will be enough to remove whatever is causing the issue.
The spif + SpiNNaker solution should be able to cope with the 346 x 240 retina at full resolution, especially with convolution processing done using local-only solutions as should happen in this case.
Helloworld-Le
commented
2 years ago Hi Andrew, thanks for your reply!
I have examined changing drop_late_spikes = False to True didn't change the error report.
Power off and power on again the devices also didn't make a difference.
In terms of capacity and resolution, for now, the testing example might be enough, but in my network model, I have 3 more layers with 10k neurons and STDP synapses. In that case, I might want to reduce the resolution and save some load from this input stage?
rowleya
commented
2 years ago I have examined changing drop_late_spikes = False to True didn't change the error report.
I wouldn't change that anyway for this case; you definitely want that set to False when using a retina.
Power off and power on again the devices also didn't make a difference.
If you are not getting any spikes with spif connected and the script you had, it could indicate that something odd has happened with spif. I would expect spif to be able to handle the configuration packets, so that you are getting reports of dropped packets going into spif is a bit worrying!
In terms of capacity and resolution, for now, the testing example might be enough, but in my network model, I have 3 more layers with 10k neurons and STDP synapses. In that case, I might want to reduce the resolution and save some load from this input stage?
In theory you could reduce the resolution at the input stage and save some neurons and processing later on. The way that SpiNNaker works is that the computational load is split up, so more layers are in theory fine, but then it also depends on the connectivity and activity of those layers. If the connection are convolutional in nature, or at least connected in 2D in a similar way to a convolution, we might just about filter enough between the populations, but if there is general random connectivity, they may or may not cope with the activity between layers. Unfortunately it is hard to predict without trying!
Note that nothing currently supports STDP in convolutional layers in any special way; they should be supported in general as the software lets you create them, but they won't then benefit from the convolutional management code currently implemented, which can handle much higher numbers of neurons per core. Additionally, the STDP weights will not be updated in the kernel, but will be updated in each of the individual connections. If you are interested in STDP where convolutional kernel weights are updated, that may be possible, but would require additional implementation.
Hi Andrew, @rowleya
I am testing live input and output, by having a SPIF board connected between a DVS camera and a Spinn-5 local board. This interface has been working fine, then when I add one more neuron to do live voltage output as:
I got the errors showing below:
My code for live voltage is from this tutorial (https://spinn-20.cs.man.ac.uk/user/helloworldle/notebooks/02.LiveInputAndOutput/LiveInputAndOutput.ipynb) and it is running correctly on itself. When I just added the code to my DVS-spif-SpiNNaker code, not even trying to connect the input to the output. I got the same error as showing above. Do you have any idea how this is caused?