class Graph
| Info | Value |
|---|---|
| Package | mvnc |
| Module | mvncapi |
| Version | 2.0 |
| See also | GraphOption, Device, Fifo, Graph.allocate(), Graph.allocate_with_fifos() |
Overview
The Graph class represents a neural network graph and provides methods to perform inferences.
Create and allocate a Graph for each network graph file. A Device can have more than one Graph allocated to it, but each Graph can only be allocated to a single Device.
Initialization
Create a Graph instance:
graph = mvncapi.Graph(name)
| Parameter | Type | Description |
|---|---|---|
| name | str | A name for the graph; this can be anything you like up to mvncapi.MAX_NAME_SIZE characters, or just an empty string. The name can be retrieved later with Graph.get_option(). |
When the Graph has been successfully created, the GraphState will be CREATED.
Methods
| Method | Description |
|---|---|
| allocate | Allocate a Graph to a device. |
| allocate_with_fifos | Allocate a graph to a device and create and allocate associated Fifos. |
| destroy | Deallocate a graph that was allocated with Graph.allocate(). |
| get_option | Get the value of a graph option. See GraphOption. |
| queue_inference | Queue an inference to be processed by the graph. |
| queue_inference_with_fifo_elem | Write input tensor data to an input Fifo buffer and then trigger graph execution. |
| set_option | Set the value of a graph option. See GraphOption. |
Typical Usage
- Create a Graph instance.
- Create input and output Fifo queues and allocate the graph for a neural compute device with Graph.allocate_with_fifos().
- Alternatively, just allocate the Graph with Graph.allocate() and handle Fifo creation and allocation separately.
- Write an image to the input Fifo and queue an inference with Graph.queue_inference_with_fifo_elem().
- You can also just queue an inference from an input Fifo that already contains elements with Graph.queue_inference() instead.
- When finished, destroy the Graph and free associated resources with Graph.destroy().
See the Python API Overview for more information about typical API usage.
Examples
from mvnc import mvncapi
#
# Create and open a Device...
#
# Create a Graph
graph = mvncapi.Graph('graph1')
# Read a compiled network graph from file (set the graph_filepath correctly for your graph file)
graph_filepath = './graph'
with open(graph_filepath, 'rb') as f:
graph_buffer = f.read()
# Allocate the graph on the device and create input and output Fifos
input_fifo, output_fifo = graph.allocate_with_fifos(device, graph_buffer)
#
# Pre-procces your input tensor...
#
# Write the tensor to the input Fifo and queue an inference
graph.queue_inference_with_fifo_elem(input_fifo, output_fifo, input_tensor, None, 'object1')
#
# Read the output from the output_fifo and use it as needed...
#
# Clean up
input_fifo.destroy()
output_fifo.destroy()
graph.destroy()
device.close()
device.destroy()