samna.dynapcnn.event

Contents

samna.dynapcnn.event#

Submodules#

Module content#

class samna.dynapcnn.event.BiasValue#

Bases: pybind11_object

A response to a bias value read request.

property bias#

The queried bias value.

from_json(self: samna.dynapcnn.event.BiasValue, arg0: str) None#
property kill#

If True then this bias value is disabled.

property monitor_tag#

The monitor_tag of the read request event that generated this read response.

property timestamp#

timestamp at which the bias read reponse event exited the chip.

to_json(self: samna.dynapcnn.event.BiasValue) str#
class samna.dynapcnn.event.DvsEvent#

Bases: pybind11_object

Represents events that DYNAP-CNN will understand as a DVS event.

On DYNAP-CNN these events can only be sent to the DVS preprocessing layer.

from_json(self: samna.dynapcnn.event.DvsEvent, arg0: str) None#
property layer#

Ignored on DYNAP-CNN

property p#

Polarity

property timestamp#
to_json(self: samna.dynapcnn.event.DvsEvent) str#
property x#

X location

property y#

Y location

class samna.dynapcnn.event.InputInterfaceEvent#

Bases: pybind11_object

InputInterface events are returned from the DYNAP-CNN when echoing is enabled.

The DYNAP-CNN has an asynchronous bus interface which may be configured to echo events received. If echoing is enabled every input event will generate two InputInterface events.

This is a low-level event type which is used for debugging.

from_json(self: samna.dynapcnn.event.InputInterfaceEvent, arg0: str) None#
property msp#

True if this event is a most significant word part, false otherwise.

property timestamp#

Time in microseconds when the event was emitted.

to_json(self: samna.dynapcnn.event.InputInterfaceEvent) str#
property value#

Encoded event value.

class samna.dynapcnn.event.MemoryValue#

Bases: pybind11_object

A response to a memory value read request.

This event is used for low level operations and should not be needed under normal conditions.

from_json(self: samna.dynapcnn.event.MemoryValue, arg0: str) None#
to_json(self: samna.dynapcnn.event.MemoryValue) str#
class samna.dynapcnn.event.NeuronValue#

Bases: pybind11_object

A response to a neuron value read request.

from_json(self: samna.dynapcnn.event.NeuronValue, arg0: str) None#
property kill#

If True then this neuron is disabled.

property monitor_tag#

The monitor_tag of the read request event that generated this read response.

property neuron_state#

The membrane potential (state) of the queried neuron.

property timestamp#

timestamp at which the read reponse event exited the chip.

to_json(self: samna.dynapcnn.event.NeuronValue) str#
class samna.dynapcnn.event.ReadBiasValue#

Bases: pybind11_object

An event that will create a read request for a bias value.

The tag property will be included in the BiasValue event that will be generated as a response. This is useful because DYNAP-CNN does not guarantee that read requests will be responded to in the order they were received.

property address#

The address of the bias to read from in its layer.

from_json(self: samna.dynapcnn.event.ReadBiasValue, arg0: str) None#
property layer#

The layer containing the bias to be read from

property monitor_tag#

The tag will be used for the read response event

to_json(self: samna.dynapcnn.event.ReadBiasValue) str#
class samna.dynapcnn.event.ReadMemoryValue#

Bases: pybind11_object

An event that will create a memory read request.

This event is used for low level operations and should not be needed under normal conditions.

from_json(self: samna.dynapcnn.event.ReadMemoryValue, arg0: str) None#
to_json(self: samna.dynapcnn.event.ReadMemoryValue) str#
class samna.dynapcnn.event.ReadNeuronValue#

Bases: pybind11_object

An event that will create a read request for a neuron state value.

The tag property will be included in the NeuronValue event that will be generated as a response. This is useful because DYNAP-CNN does not guarantee that read requests will be responded to in the order they were received.

property address#

The address of the neuron to read from in its layer.

from_json(self: samna.dynapcnn.event.ReadNeuronValue, arg0: str) None#
property layer#

The layer containing the neuron to be read from

property monitor_tag#

The tag will be used for the read response event

to_json(self: samna.dynapcnn.event.ReadNeuronValue) str#
class samna.dynapcnn.event.ReadRegisterValue#

Bases: pybind11_object

An event that will create a register read request.

This event is used for low level operations and should not be needed under normal conditions.

from_json(self: samna.dynapcnn.event.ReadRegisterValue, arg0: str) None#
to_json(self: samna.dynapcnn.event.ReadRegisterValue) str#
class samna.dynapcnn.event.ReadWeightValue#

Bases: pybind11_object

An event that will create a read request for a weight value.

The tag property will be included in the WeightValue event that will be generated as a response. This is useful because DYNAP-CNN does not guarantee that read requests will be responded to in the order they were received.

property address#

The address of the weight to read from in its layer.

from_json(self: samna.dynapcnn.event.ReadWeightValue, arg0: str) None#
property layer#

The layer containing the weight to be read from

property monitor_tag#

The tag will be used for the read response event

to_json(self: samna.dynapcnn.event.ReadWeightValue) str#
class samna.dynapcnn.event.RegisterValue#

Bases: pybind11_object

A response to a register value read request.

This event is used for low level operations and should not be needed under normal conditions.

from_json(self: samna.dynapcnn.event.RegisterValue, arg0: str) None#
to_json(self: samna.dynapcnn.event.RegisterValue) str#
class samna.dynapcnn.event.RouterEvent#

Bases: pybind11_object

A DYNAP-CNN event type that will be injected into the on-chip event router.

The layer, feature, x, y properties determine the neuron that will receive the event.

RouterEvents can only be injected. When DYNAP-CNN produces events they will be of type Spike. RouterEvents also do not contain a timestamp. If timestamps are needed, use the Spike event type.

property feature#

The feature that will receive this event.

from_json(self: samna.dynapcnn.event.RouterEvent, arg0: str) None#
property layer#

The layer that will receive this event.

to_json(self: samna.dynapcnn.event.RouterEvent) str#
property x#

The x coordinate that will receive this event.

property y#

The y coordinate that will receive this event.

class samna.dynapcnn.event.Spike#

Bases: pybind11_object

Spikes are events that can be consumed and produced by DYNAP-CNN.

When sending spike events to DYNAP-CNN, the properties will be interpreted as the destination layer, feature, x, y that should receive the event. The timestamp property will be used to inject the event into DYNAP-CNN when the external event sequencer timestamp is equal to or greater than the event timestamp.

When receing spike events from DYNAP-CNN the properties should be interpreted as the layer, feature, x, y neuron that generated the spike. The timestamp parameter is the timestamp at which DYNAP-CNN sent out the event.

property feature#

Target/producing feature.

from_json(self: samna.dynapcnn.event.Spike, arg0: str) None#
property layer#

Target/producing layer.

property timestamp#

When receiving a Spike event from a chip the timestamp will be populated with the time at which the event leave the chip.

When sending a Spike event to a chip the timestamp will be used to control when the event will be applied to the chip.

to_json(self: samna.dynapcnn.event.Spike) str#
property x#

Target/producing x coordinate.

property y#

Target/producing y coordinate.

class samna.dynapcnn.event.WeightValue#

Bases: pybind11_object

A response to a kernel weight value read request.

from_json(self: samna.dynapcnn.event.WeightValue, arg0: str) None#
property kill#

If True then this weight value is disabled.

property monitor_tag#

The monitor_tag of the read request event that generated this read response.

property timestamp#

timestamp at which the weight read response exited the chip

to_json(self: samna.dynapcnn.event.WeightValue) str#
property weight#

The queried weight value

class samna.dynapcnn.event.WriteBiasValue#

Bases: pybind11_object

An event that will write a bias value.

This event is used for low level operations. Biases will be written automatically when a configuration is applied. Prefer using the configuration whenever possible.

property address#

The address of the bias that will be written to.

property bias#

The bias value to write

from_json(self: samna.dynapcnn.event.WriteBiasValue, arg0: str) None#
property kill#

If true, disables this bias from being used.

property layer#

The layer containing the bias that will be written to.

to_json(self: samna.dynapcnn.event.WriteBiasValue) str#
class samna.dynapcnn.event.WriteMemoryValue#

Bases: pybind11_object

An event that will write a memory value.

This event is used for low level operations. Memory will be written automatically when a configuration is applied. Prefer using the configuration whenever possible.

from_json(self: samna.dynapcnn.event.WriteMemoryValue, arg0: str) None#
to_json(self: samna.dynapcnn.event.WriteMemoryValue) str#
class samna.dynapcnn.event.WriteNeuronValue#

Bases: pybind11_object

An event that will write the neuron state value of a single neuron.

This event is used for low level operations. Initial neuron state values will be written automatically when a configuration is applied. Prefer using the configuration whenever possible.

property address#

The address of the neuron to be written to

from_json(self: samna.dynapcnn.event.WriteNeuronValue, arg0: str) None#
property kill#

If true, disables this neuron state value from being used

property layer#

The layer containing the neuron to be written to

property neuron_state#

The value which will be written to the neuron state address

to_json(self: samna.dynapcnn.event.WriteNeuronValue) str#
class samna.dynapcnn.event.WriteRegisterValue#

Bases: pybind11_object

An event that will write a register value.

This event is used for low level operations. Registers will be written automatically when a configuration is applied. Prefer using the configuration whenever possible.

from_json(self: samna.dynapcnn.event.WriteRegisterValue, arg0: str) None#
to_json(self: samna.dynapcnn.event.WriteRegisterValue) str#
class samna.dynapcnn.event.WriteWeightValue#

Bases: pybind11_object

An event that will write a kernel weight.

This event is used for low level operations. Kernel weights will be written automatically when a configuration is applied. Prefer using the configuration whenever possible.

property address#

The address of the kernel that will be written to.

from_json(self: samna.dynapcnn.event.WriteWeightValue, arg0: str) None#
property kill#

If true, disables this kernel weight from being used.

property layer#

The layer containing the kernel weight that will be written to.

to_json(self: samna.dynapcnn.event.WriteWeightValue) str#
property weight#

The weight value to write

samna.dynapcnn.event.decode_input_event(arg0: int, arg1: bool) samna.dynapcnn.event.Spike | samna.dynapcnn.event.RouterEvent | samna.dynapcnn.event.KillSensorPixel | samna.dynapcnn.event.ResetSensorPixel | samna.dynapcnn.event.WriteNeuronValue | samna.dynapcnn.event.ReadNeuronValue | samna.dynapcnn.event.WriteWeightValue | samna.dynapcnn.event.ReadWeightValue | samna.dynapcnn.event.WriteBiasValue | samna.dynapcnn.event.ReadBiasValue | samna.dynapcnn.event.WriteRegisterValue | samna.dynapcnn.event.ReadRegisterValue | samna.dynapcnn.event.WriteMemoryValue | samna.dynapcnn.event.ReadMemoryValue | samna.dynapcnn.event.ReadProbe#

Decodes an input event.

Parameters:
  • rawEvent (int) – The event to decode.

  • hasTimestamp (bool) – Whether the event has a timestamp.

Returns:

The decoded input event.

samna.dynapcnn.event.decode_output_event(arg0: int) samna.dynapcnn.event.Spike | samna.dynapcnn.event.DvsEvent | samna.dynapcnn.event.InputInterfaceEvent | samna.dynapcnn.event.NeuronValue | samna.dynapcnn.event.BiasValue | samna.dynapcnn.event.WeightValue | samna.dynapcnn.event.RegisterValue | samna.dynapcnn.event.MemoryValue | samna.dynapcnn.event.BistValue | samna.dynapcnn.event.ProbeValue | samna.dynapcnn.event.ReadoutValue#

Decodes an output event.

Parameters:

value (int) – The event to decode.

Returns:

The decoded output event.

samna.dynapcnn.event.dynapcnn_configuration_to_event(arg0: samna.dynapcnn.configuration.DynapcnnConfiguration) List[samna.dynapcnn.event.Spike | samna.dynapcnn.event.RouterEvent | samna.dynapcnn.event.KillSensorPixel | samna.dynapcnn.event.ResetSensorPixel | samna.dynapcnn.event.WriteNeuronValue | samna.dynapcnn.event.ReadNeuronValue | samna.dynapcnn.event.WriteWeightValue | samna.dynapcnn.event.ReadWeightValue | samna.dynapcnn.event.WriteBiasValue | samna.dynapcnn.event.ReadBiasValue | samna.dynapcnn.event.WriteRegisterValue | samna.dynapcnn.event.ReadRegisterValue | samna.dynapcnn.event.WriteMemoryValue | samna.dynapcnn.event.ReadMemoryValue | samna.dynapcnn.event.ReadProbe]#

Get the input events for the given configuration.

Parameters:

config (dynapcnn.configuration.DynapcnnConfiguration) – The configuration to turn into events.

Returns:

A list of input events.

samna.dynapcnn.event.encode_input_interface_event(arg0: samna.dynapcnn.event.Spike | samna.dynapcnn.event.RouterEvent | samna.dynapcnn.event.KillSensorPixel | samna.dynapcnn.event.ResetSensorPixel | samna.dynapcnn.event.WriteNeuronValue | samna.dynapcnn.event.ReadNeuronValue | samna.dynapcnn.event.WriteWeightValue | samna.dynapcnn.event.ReadWeightValue | samna.dynapcnn.event.WriteBiasValue | samna.dynapcnn.event.ReadBiasValue | samna.dynapcnn.event.WriteRegisterValue | samna.dynapcnn.event.ReadRegisterValue | samna.dynapcnn.event.WriteMemoryValue | samna.dynapcnn.event.ReadMemoryValue | samna.dynapcnn.event.ReadProbe) int#

Encodes an input interface event.

Parameters:

event (dynapcnn.event.InputEvent) – The event to encode.

Returns:

The encoded event.