morpheus.stages.inference.inference_stage.InferenceStage

class InferenceStage(c)[source]

Bases: morpheus.pipeline.multi_message_stage.MultiMessageStage

This class serves as the base for any inference stage. Inference stages operate differently than other stages due to the fact that they operate in a separate thread and have their own batch size which is separate from the pipeline batch size. Processing the inference work in a separate thread is necessary to support inference types that may require exclusive use of a single thread (i.e., TensorRT) without blocking the main thread.

Changing batch sizes for the inference stage requires breaking messages into smaller slices, running inference on the smaller slices, then recombining the inference output into the original batch size. This inference base class handles breaking and recombining batches and queing the inference work to be processed on another thread.

Inference stages that derive from this class need to implement the _get_inference_worker method which returns an implementation of the InferenceWorker class. Your InferenceWorker class must implement the process and calc_output_dims methods. The process methods is where you provide implementation details on how to perform inference with the MultiInferenceMessage batch. The worker uses the calc_output_dims to calculate the output dimensions of the pipeline batch that inference batch results are appended to.

To add a C++ implementation for processing inference requests, you must implement the _get_cpp_inference_node method and implement supports_cpp_node in your worker to return True. Your pipeline can then use your C++ implementation by setting use_cpp to True in your pipeline configuration. See developer documentation for more details.

Parameters
cmorpheus.config.Config

Pipeline configuration instance.

Attributes
has_multi_input_ports

Indicates if this stage has multiple input ports.

has_multi_output_ports

Indicates if this stage has multiple output ports.

input_ports

Input ports to this stage.

is_built

Indicates if this stage has been built.

name

The name of the stage.

output_ports

Output ports from this stage.

unique_name

Unique name of stage.

Methods

accepted_types()

Accepted input types to this stage.

build(builder[, do_propagate])

Build this stage.

can_build([check_ports])

Determines if all inputs have been built allowing this node to be built.

get_all_input_stages()

Get all input stages to this stage.

get_all_inputs()

Get all input senders to this stage.

get_all_output_stages()

Get all output stages from this stage.

get_all_outputs()

Get all output receivers from this stage.

get_needed_columns()

Stages which need to have columns inserted into the dataframe, should populate the self._needed_columns dictionary with mapping of column names to morpheus.common.TypeId.

join()

On all inference worker threads, this function applies join.

on_start()

This function can be overridden to add usecase-specific implementation at the start of any stage in the pipeline.

start_async()

This function is called along with on_start during stage initialization.

stop()

Stops the inference workers and closes the inference queue.

supports_cpp_node()

Specifies whether this Stage is capable of creating C++ nodes.

_build(builder, in_ports_streams)[source]

This function is responsible for constructing this stage’s internal mrc.SegmentObject object. The input of this function contains the returned value from the upstream stage.

The input values are the mrc.Builder for this stage and a StreamPair tuple which contain the input mrc.SegmentObject object and the message data type.

Parameters
buildermrc.Builder

mrc.Builder object for the pipeline. This should be used to construct/attach the internal mrc.SegmentObject.

in_ports_streamsmorpheus.pipeline.pipeline.StreamPair

List of tuples containing the input mrc.SegmentObject object and the message data type.

Returns
typing.List[morpheus.pipeline.pipeline.StreamPair]

List of tuples containing the output mrc.SegmentObject object from this stage and the message data type.

abstract _get_inference_worker(inf_queue)[source]

Returns the main inference worker which manages requests possibly in another thread depending on which mode the pipeline is currently operating in.

Parameters
inf_queuemorpheus.utils.producer_consumer_queue.ProducerConsumerQueue

Inference request queue.

Returns
InferenceWorker

Inference worker implementation for stage.

accepted_types()[source]

Accepted input types to this stage.

Returns
typing.Tuple

Tuple of input types.

build(builder, do_propagate=True)[source]

Build this stage.

Parameters
buildermrc.Builder

MRC segment for this stage.

do_propagatebool, optional

Whether to propagate to build output stages, by default True.

can_build(check_ports=False)[source]

Determines if all inputs have been built allowing this node to be built.

Parameters
check_portsbool, optional

Check if we can build based on the input ports, by default False.

Returns
bool

True if we can build, False otherwise.

get_all_input_stages()[source]

Get all input stages to this stage.

Returns
typing.List[morpheus.pipeline.pipeline.StreamWrapper]

All input stages.

get_all_inputs()[source]

Get all input senders to this stage.

Returns
typing.List[morpheus.pipeline.pipeline.Sender]

All input senders.

get_all_output_stages()[source]

Get all output stages from this stage.

Returns
typing.List[morpheus.pipeline.pipeline.StreamWrapper]

All output stages.

get_all_outputs()[source]

Get all output receivers from this stage.

Returns
typing.List[morpheus.pipeline.pipeline.Receiver]

All output receivers.

get_needed_columns()[source]

Stages which need to have columns inserted into the dataframe, should populate the self._needed_columns dictionary with mapping of column names to morpheus.common.TypeId. This will ensure that the columns are allocated and populated with null values.

property has_multi_input_ports: bool

Indicates if this stage has multiple input ports.

Returns
bool

True if stage has multiple input ports, False otherwise.

property has_multi_output_ports: bool

Indicates if this stage has multiple output ports.

Returns
bool

True if stage has multiple output ports, False otherwise.

property input_ports: List[morpheus.pipeline.receiver.Receiver]

Input ports to this stage.

Returns
typing.List[morpheus.pipeline.pipeline.Receiver]

Input ports to this stage.

property is_built: bool

Indicates if this stage has been built.

Returns
bool

True if stage is built, False otherwise.

async join()[source]

On all inference worker threads, this function applies join.

property name: str

The name of the stage. Used in logging. Each derived class should override this property with a unique name.

Returns
str

Name of a stage.

on_start()[source]

This function can be overridden to add usecase-specific implementation at the start of any stage in the pipeline.

property output_ports: List[morpheus.pipeline.sender.Sender]

Output ports from this stage.

Returns
typing.List[morpheus.pipeline.pipeline.Sender]

Output ports from this stage.

async start_async()[source]

This function is called along with on_start during stage initialization. Allows stages to utilize the asyncio loop if needed.

stop()[source]

Stops the inference workers and closes the inference queue.

supports_cpp_node()[source]

Specifies whether this Stage is capable of creating C++ nodes. During the build phase, this value will be combined with CppConfig.get_should_use_cpp() to determine whether or not a C++ node is created. This is an instance method to allow runtime decisions and derived classes to override base implementations.

property unique_name: str

Unique name of stage. Generated by appending stage id to stage name.

Returns
str

Unique name of stage.

© Copyright 2023, NVIDIA. Last updated on Apr 11, 2023.