aidge-model-explorer 0.9.1.post2

Aidge module for model explorer: https://github.com/google-ai-edge/model-explorer.

Aidge Model Explorer

Plugin of the framework Google Model Explorer to visualize Aidge graphs. For more information on how to use Model Explorer, you can refer yourself to the official documentation of the framework.

Quick start

System Requirements

• python >= 3.10
• aidge_core

pip install aidge-model-explorer

🛠 Build from Source

Prerequisite (in addition to previous one):

• Please review the global installation instructions before proceeding.
• If using a virtual environment, use the same one for all Aidge modules.

1. Python installation using setup scripts

Environment	Python Development
Windows	.\setup.ps1 -Modules model_explorer -Tests
Unix	./setup.sh -m model_explorer --tests

[!TIP] Use Get-Help setup.ps1 (Windows) or ./setup.sh -h (Unix) for full documentation.

2. Python Installation (pip)

Run these commands from the aidge_model_explorer/ directory:

#fStandard install
pip install . -v

# Install with testing dependencies
pip install .[test] -v && pytest

Editable Install (Experimental)

Use this for real-time development without re-installing.

pip install --no-build-isolation -ve . --config-settings=editable.rebuild=true -Cbuild-dir=build

🚀 Features

Visualize ONNX file

You can use directly the model-explorer CLI with the extension aidge_model_explorer to visualize an ONNX graph like so:

model-explorer aidge_mobilenetV2.onnx --extensions=aidge_model_explorer

This will allow you to see how Aidge will import your ONNX graph, looking at potential MetaOperator created. Also GenericOperator have a specific background allowing to easily see which operator is not supported natively by Aidge.

Visualize wrapper method

This function wrap every call to the model explorer API, allowing to only call one line to visualize your graph.

Note: This function will automatically embed the visualization in a Jupyter Notebook cell if ran in a notebook, otherwise it will simply open a page in your browser.

import aidge_core
import aidge_model_explorer

lstm = aidge_core.LSTM(in_channels=4, hidden_channels=8, seq_length=5)
model = aidge_core.get_connected_graph_view(lstm)

aidge_model_explorer.visualize(model, "LSTM")

:warning: When using model-explorer from a distant server, port forwarding is not always automatic. Make sure to open the right port. (You can force the port to use by providing the argument port)

Visualize from config

You can also use the model_explorer API by using a aidge_model_explorer.Config() which is an overridden class enabling full compatibility with the model_explorer API.

import aidge_core
import aidge_model_explorer
import model_explorer


lstm = aidge_core.LSTM(in_channels=4, hidden_channels=8, seq_length=5)
lstm_model = aidge_core.get_connected_graph_view(lstm)

config = aidge_model_explorer.config()
config.add_graphview(lstm_model, "lstm")
model_explorer.visualize_from_config(config)

Synchronize graphs with node ids

Model explorer support an option to synchronize two graphs using node id. You can use this functionality to visualize a graph before and after graph modifications:

import aidge_core
import aidge_model_explorer
import model_explorer


lstm = aidge_core.LSTM(in_channels=4, hidden_channels=8, seq_length=5)
lstm_model = aidge_core.get_connected_graph_view(lstm)


config = aidge_model_explorer.config()
# Note: need to add graphview before doing any manipulation to keep the ids
config.add_graphview(lstm_model, "lstm")

aidge_core.expand_metaops(lstm_model)
config.add_graphview(lstm_model, "lstm_expanded")

model_explorer.visualize_from_config(config)

Once the GUI is opened, you can then explorer the two models like this:

Note: Screen recording taken using v0.1.5 of model explorer.

Add attributes & metadata

import aidge_core
import aidge_onnx
import aidge_model_explorer
import model_explorer


model = aidge_onnx.load_onnx("aidge_mobilenetV2.onnx")

conv_config = aidge_model_explorer.ConverterConfig()
conv_config.add_attribute("isConv",
                          lambda node: "" if (node.type() == "Conv2D" or node.type() == "ConvDepthWise2D") else None)

conv_config.add_output_metadata("mean",
                          lambda _, tensor: str(tensor.mean()))

config = aidge_model_explorer.config()
config.add_graphview(model, "model1", conv_config)

model_explorer.visualize_from_config(config)

Once the GUI is opened, you can then play with the node styler to easily see the nodes that are convolutions.

Note: Screen recording taken using v0.1.5 of model explorer.

Update node style

If you want to apply a specific style, you can add the following attributes to your nodes:

• model_explorer:bg_color: Changes the background color of the node. The value must be a string representing a valid color in hexadecimal format (e.g., "#ff0800" for red). The string should follow the standard 6-digit RGB hex code format, beginning with a #.
• model_explorer:border_color: Changes the border color of the node. The value must be a string representing a valid color in hexadecimal format (e.g., "#ff0800" for red). The string should follow the standard 6-digit RGB hex code format, beginning with a #.
• model_explorer:h_border_color: Changes the border color on hover of the node. The value must be a string representing a valid color in hexadecimal format (e.g., "#ff0800" for red). The string should follow the standard 6-digit RGB hex code format, beginning with a #.

Example:

import aidge_core
import aidge_model_explorer
from aidge_model_explorer.visualize import visualize_from_config

node = aidge_core.ReLU(name="Colorized Node!")
node.attributes().set_attr(name="model_explorer:bg_color", value="#ff0800")
node.attributes().set_attr(name="model_explorer:border_color", value="#00ff4c")
node.attributes().set_attr(name="model_explorer:h_border_color", value="#1100ff")

g = aidge_core.sequential([node])
conf = aidge_model_explorer.config()

conf.add_graphview(g, "test_color")
visualize_from_config(conf)

📊 Default visualizations

Aidge model explorer propose some default visualization in order to let the user get started quickly.

Has an implementation

Add an attribute to every node for which we cannot find an implementation to use (for example because input/output datatype does not match). This attribute is named fail_best_match and contains the description of the current node specs and the specs for which Aidge has a registered implementation.

To use this function use:

my_graph: aidge_core.GraphView
aidge_model_explorer.has_best_match(my_graph, "cpu")