Update Model Parameters API Examples¶
This demonstrates how to use the update_model_parameters API.
Refer to the Model Parameters guide for more details.
NOTES:
Click here:
to run this example interactively in your browserRefer to the Notebook Examples Guide for how to run this example locally in VSCode
Install MLTK Python Package¶
# Install the MLTK Python package (if necessary)
!pip install --upgrade silabs-mltk
Import Python Packages¶
# Import the necessary MLTK APIs
from mltk.core import update_model_parameters, summarize_model
Example 1: Update model specification¶
The most common use case of the update_model_parameters API is:
Fully train a model
Later modify the model specification script with additional parameters
Run the update_model_parameters API to update the
.tflitemodel file in the model archive.
In this example, it’s assumed that the MltkModel.model_parameters settings in the tflite_micro_speech model specification script have been modified after the model have been trained.
my_model.model_parameters['average_window_duration_ms'] = 1000
my_model.model_parameters['detection_threshold'] = 185
my_model.model_parameters['suppression_ms'] = 1500
my_model.model_parameters['minimum_count'] = 3
my_model.model_parameters['volume_db'] = 5.0
my_model.model_parameters['latency_ms'] = 0
my_model.model_parameters['log_level'] = 'info'
After this API completes, the tflite_micro_speech.mltk.zip model archive is updated with a new tflite_micro_speech.tflite model file.
Note that only the parameters in the .tflite’s metadata section are modified. The model weights and layers are untouched.
# Update the model parameters
update_model_parameters('tflite_micro_speech')
# Generate a summary of the updated model with new parameters
print(summarize_model('tflite_micro_speech'))
Using Tensorflow-Lite Micro version: b13b48c (2022-06-08)
Searching for optimal runtime memory size ...
Determined optimal runtime memory size to be 9600
Updating c:/users/reed/workspace/silabs/mltk/mltk/models/tflite_micro/tflite_micro_speech.mltk.zip
Model: "tflite_micro_speech"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 25, 20, 8) 648
batch_normalization (BatchN (None, 25, 20, 8) 32
ormalization)
activation (Activation) (None, 25, 20, 8) 0
dropout (Dropout) (None, 25, 20, 8) 0
flatten (Flatten) (None, 4000) 0
dense (Dense) (None, 4) 16004
=================================================================
Total params: 16,684
Trainable params: 16,668
Non-trainable params: 16
_________________________________________________________________
Total MACs: 336.000 k
Total OPs: 684.004 k
Name: tflite_micro_speech
Version: 1
Description: TFLite-Micro speech
Classes: yes, no, _unknown_, _silence_
average_window_duration_ms: 1000
detection_threshold: 185
suppression_ms: 750
minimum_count: 3
volume_gain: 2
latency_ms: 100
verbose_model_output_logs: False
Example 2: Update with argument¶
The update_model_parameters API also works with .tflite model files generated outside of the MLTK.
In this mode, model parameters can be supplied via API argument, e.g.:
import os
import tempfile
import urllib
import shutil
# Use .tflite mode found here:
# https://github.com/mlcommons/tiny/tree/master/benchmark/training/keyword_spotting/trained_models
# NOTE: Update this URL to point to your model if necessary
TFLITE_MODEL_URL = 'https://github.com/mlcommons/tiny/raw/master/benchmark/training/keyword_spotting/trained_models/kws_ref_model.tflite'
# Download the .tflite file and save to the temp dir
external_tflite_path = os.path.normpath(f'{tempfile.gettempdir()}/kws_ref_model.tflite')
with open(external_tflite_path, 'wb') as dst:
with urllib.request.urlopen(TFLITE_MODEL_URL) as src:
shutil.copyfileobj(src, dst)
# Set the parameters in a Python dictionary
parameters = {
"volume": 10.0,
"log_level": "debug",
"threshold": 43
}
# Update the model parameters
update_model_parameters(external_tflite_path, params=parameters)
# Generate a summary of the updated model with new parameters
print(summarize_model(external_tflite_path))
Using Tensorflow-Lite Micro version: b13b48c (2022-06-08)
Searching for optimal runtime memory size ...
Determined optimal runtime memory size to be 25472
+-------+-------------------+----------------+----------------+-------------------------------------------------------+
| Index | OpCode | Input(s) | Output(s) | Config |
+-------+-------------------+----------------+----------------+-------------------------------------------------------+
| 0 | conv_2d | 49x10x1 (int8) | 25x5x64 (int8) | Padding:same stride:2x2 activation:relu |
| | | 10x4x1 (int8) | | |
| | | 64 (int32) | | |
| 1 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 2 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 3 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 4 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 5 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 6 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 7 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 8 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 9 | average_pool_2d | 25x5x64 (int8) | 1x1x64 (int8) | Padding:valid stride:5x25 filter:5x25 activation:none |
| 10 | reshape | 1x1x64 (int8) | 64 (int8) | Type=none |
| | | 2 (int32) | | |
| 11 | fully_connected | 64 (int8) | 12 (int8) | Activation:none |
| | | 64 (int8) | | |
| | | 12 (int32) | | |
| 12 | softmax | 12 (int8) | 12 (int8) | Type=softmaxoptions |
+-------+-------------------+----------------+----------------+-------------------------------------------------------+
Total MACs: 2.657 M
Total OPs: 5.394 M
Name: summarize_model
Version: 1
Description: Generated by Silicon Lab's MLTK Python package
Runtime memory size (RAM): 24.384 k
hash: a4fadcb07cc13bf4ce4249e7349cadf3
date: 2022-10-06T16:36:18.836Z
volume: 10.0
log_level: debug
threshold: 43
.tflite file size: 54.1kB
Example 3: Return TfliteModel instance¶
Rather than update the given model, the update_model_parameters API can also return a TfliteModel instance with the updated parameters.
This is done by specifying the output='tflite_model' API argument:
# Set the parameters in a Python dictionary
parameters = {
"volume": 10.0,
"log_level": "debug",
"threshold": 43
}
# Generate a TfliteModel instance with the given parameters
# NOTE: The input external_tflite_path file is NOT modified
tflite_model = update_model_parameters(external_tflite_path, params=parameters, output='tflite_model')
# Generate a summary of the returned TfliteModel instance
print(summarize_model(tflite_model))
Using Tensorflow-Lite Micro version: b13b48c (2022-06-08)
Searching for optimal runtime memory size ...
Determined optimal runtime memory size to be 25472
+-------+-------------------+----------------+----------------+-------------------------------------------------------+
| Index | OpCode | Input(s) | Output(s) | Config |
+-------+-------------------+----------------+----------------+-------------------------------------------------------+
| 0 | conv_2d | 49x10x1 (int8) | 25x5x64 (int8) | Padding:same stride:2x2 activation:relu |
| | | 10x4x1 (int8) | | |
| | | 64 (int32) | | |
| 1 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 2 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 3 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 4 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 5 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 6 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 7 | depthwise_conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Multiplier:1 padding:same stride:1x1 activation:relu |
| | | 3x3x64 (int8) | | |
| | | 64 (int32) | | |
| 8 | conv_2d | 25x5x64 (int8) | 25x5x64 (int8) | Padding:same stride:1x1 activation:relu |
| | | 1x1x64 (int8) | | |
| | | 64 (int32) | | |
| 9 | average_pool_2d | 25x5x64 (int8) | 1x1x64 (int8) | Padding:valid stride:5x25 filter:5x25 activation:none |
| 10 | reshape | 1x1x64 (int8) | 64 (int8) | Type=none |
| | | 2 (int32) | | |
| 11 | fully_connected | 64 (int8) | 12 (int8) | Activation:none |
| | | 64 (int8) | | |
| | | 12 (int32) | | |
| 12 | softmax | 12 (int8) | 12 (int8) | Type=softmaxoptions |
+-------+-------------------+----------------+----------------+-------------------------------------------------------+
Total MACs: 2.657 M
Total OPs: 5.394 M
Name: summarize_model
Version: 1
Description: Generated by Silicon Lab's MLTK Python package
Runtime memory size (RAM): 24.384 k
hash: a4fadcb07cc13bf4ce4249e7349cadf3
date: 2022-10-06T16:36:26.831Z
volume: 10.0
log_level: debug
threshold: 43
.tflite file size: 54.1kB