ContextQMD
Back to Tensorflow

tf.lite.TFLiteConverter

tensorflow/lite/g3doc/api_docs/python/tf/lite/TFLiteConverter.md

2.21.013.8 KB
Original Source

page_type: reference description: Converts a TensorFlow model into TensorFlow Lite model.

<link rel="stylesheet" href="/site-assets/css/style.css"> <!-- DO NOT EDIT! Automatically generated file. --> <div itemscope itemtype="http://developers.google.com/ReferenceObject"> <meta itemprop="name" content="tf.lite.TFLiteConverter" /> <meta itemprop="path" content="Stable" /> <meta itemprop="property" content="__init__"/> <meta itemprop="property" content="convert"/> <meta itemprop="property" content="experimental_from_jax"/> <meta itemprop="property" content="from_concrete_functions"/> <meta itemprop="property" content="from_keras_model"/> <meta itemprop="property" content="from_saved_model"/> </div>

tf.lite.TFLiteConverter

<!-- Insert buttons and diff --> <table class="tfo-notebook-buttons tfo-api nocontent" align="left"> <td> <a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v2.11.0/tensorflow/lite/python/lite.py#L1628-L1866"> 
View source on GitHub
</a> </td> </table>

Converts a TensorFlow model into TensorFlow Lite model.

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>tf.lite.TFLiteConverter( funcs, trackable_obj=None ) </code></pre> <h3>Used in the notebooks</h3> <table class="vertical-rules"> <thead> <tr> <th>Used in the guide</th> <th>Used in the tutorials</th> </tr> </thead> <tbody> <tr> <td> <ul> <li><a href="https://www.tensorflow.org/lite/guide/model_analyzer">TensorFlow Lite Model Analyzer</a></li> <li><a href="https://www.tensorflow.org/lite/guide/signatures">Signatures in TensorFlow Lite</a></li> <li><a href="https://www.tensorflow.org/lite/guide/authoring">TFLite Authoring Tool</a></li> <li><a href="https://www.tensorflow.org/guide/migrate/tflite">Migrating your TFLite code to TF2</a></li> <li><a href="https://www.tensorflow.org/model_optimization/guide/combine/cqat_example">Cluster preserving quantization aware training (CQAT) Keras example</a></li> </ul> </td> <td> <ul> <li><a href="https://www.tensorflow.org/lite/performance/post_training_integer_quant">Post-training integer quantization</a></li> <li><a href="https://www.tensorflow.org/lite/performance/quantization_debugger">Inspecting Quantization Errors with Quantization Debugger</a></li> <li><a href="https://www.tensorflow.org/lite/examples/jax_conversion/overview">Jax Model Conversion For TFLite</a></li> <li><a href="https://www.tensorflow.org/lite/performance/post_training_quant">Post-training dynamic range quantization</a></li> <li><a href="https://www.tensorflow.org/lite/examples/on_device_training/overview">On-Device Training with TensorFlow Lite</a></li> </ul> </td> </tr> </tbody> </table>

Example usage:

python
# Converting a SavedModel to a TensorFlow Lite model.
  converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir)
  tflite_model = converter.convert()

# Converting a tf.Keras model to a TensorFlow Lite model.
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()

# Converting ConcreteFunctions to a TensorFlow Lite model.
converter = tf.lite.TFLiteConverter.from_concrete_functions([func], model)
tflite_model = converter.convert()

# Converting a Jax model to a TensorFlow Lite model.
converter = tf.lite.TFLiteConverter.experimental_from_jax([func], [[
    ('input1', input1), ('input2', input2)]])
tflite_model = converter.convert()
<!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2"><h2 class="add-link">Args</h2></th></tr> <tr> <td> `funcs`<a id="funcs"></a> </td> <td> List of TensorFlow ConcreteFunctions. The list should not contain duplicate elements. </td> </tr><tr> <td> `trackable_obj`<a id="trackable_obj"></a> </td> <td> tf.AutoTrackable object associated with `funcs`. A reference to this object needs to be maintained so that Variables do not get garbage collected since functions have a weak reference to Variables. This is only required when the tf.AutoTrackable object is not maintained by the user (e.g. `from_saved_model`). </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2"><h2 class="add-link">Attributes</h2></th></tr> <tr> <td> `optimizations`<a id="optimizations"></a> </td> <td> Experimental flag, subject to change. Set of optimizations to apply. e.g {tf.lite.Optimize.DEFAULT}. (default None, must be None or a set of values of type <a href="../../tf/lite/Optimize"><code>tf.lite.Optimize</code></a>) </td> </tr><tr> <td> `representative_dataset`<a id="representative_dataset"></a> </td> <td> A generator function used for integer quantization where each generated sample has the same order, type and shape as the inputs to the model. Usually, this is a small subset of a few hundred samples randomly chosen, in no particular order, from the training or evaluation dataset. This is an optional attribute, but required for full integer quantization, i.e, if <a href="https://www.tensorflow.org/api_docs/python/tf#int8"><code>tf.int8</code></a> is the only supported type in `target_spec.supported_types`. Refer to <a href="../../tf/lite/RepresentativeDataset"><code>tf.lite.RepresentativeDataset</code></a>. (default None) </td> </tr><tr> <td> `target_spec`<a id="target_spec"></a> </td> <td> Experimental flag, subject to change. Specifications of target device, including supported ops set, supported types and a set of user's defined TensorFlow operators required in the TensorFlow Lite runtime. Refer to <a href="../../tf/lite/TargetSpec"><code>tf.lite.TargetSpec</code></a>. </td> </tr><tr> <td> `inference_input_type`<a id="inference_input_type"></a> </td> <td> Data type of the input layer. Note that integer types (tf.int8 and tf.uint8) are currently only supported for post training integer quantization and quantization aware training. (default tf.float32, must be in {tf.float32, tf.int8, tf.uint8}) </td> </tr><tr> <td> `inference_output_type`<a id="inference_output_type"></a> </td> <td> Data type of the output layer. Note that integer types (tf.int8 and tf.uint8) are currently only supported for post training integer quantization and quantization aware training. (default tf.float32, must be in {tf.float32, tf.int8, tf.uint8}) </td> </tr><tr> <td> `allow_custom_ops`<a id="allow_custom_ops"></a> </td> <td> Boolean indicating whether to allow custom operations. When False, any unknown operation is an error. When True, custom ops are created for any op that is unknown. The developer needs to provide these to the TensorFlow Lite runtime with a custom resolver. (default False) </td> </tr><tr> <td> `exclude_conversion_metadata`<a id="exclude_conversion_metadata"></a> </td> <td> Whether not to embed the conversion metadata into the converted model. (default False) </td> </tr><tr> <td> `experimental_new_converter`<a id="experimental_new_converter"></a> </td> <td> Experimental flag, subject to change. Enables MLIR-based conversion. (default True) </td> </tr><tr> <td> `experimental_new_quantizer`<a id="experimental_new_quantizer"></a> </td> <td> Experimental flag, subject to change. Enables MLIR-based quantization conversion instead of Flatbuffer-based conversion. (default True) </td> </tr><tr> <td> `experimental_enable_resource_variables`<a id="experimental_enable_resource_variables"></a> </td> <td> Experimental flag, subject to change. Enables [resource variables](https://tensorflow.org/guide/migrate/tf1_vs_tf2#resourcevariables_instead_of_referencevariables) to be converted by this converter. This is only allowed if the from_saved_model interface is used. (default True) </td> </tr> </table>

Methods

<h3 id="convert"><code>convert</code></h3>

<a target="_blank" class="external" href="https://github.com/tensorflow/tensorflow/blob/v2.11.0/tensorflow/lite/python/lite.py#L1853-L1866">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>convert() </code></pre>

Converts a TensorFlow GraphDef based on instance variables.

<!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> The converted data in serialized format. </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Raises</th></tr> <tr> <td> `ValueError` </td> <td> No concrete functions is specified. Multiple concrete functions are specified. Input shape is not specified. Invalid quantization parameters. </td> </tr> </table> <h3 id="experimental_from_jax"><code>experimental_from_jax</code></h3>

<a target="_blank" class="external" href="https://github.com/tensorflow/tensorflow/blob/v2.11.0/tensorflow/lite/python/lite.py#L1830-L1850">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>@classmethod</code> <code>experimental_from_jax( serving_funcs, inputs ) </code></pre>

Creates a TFLiteConverter object from a Jax model with its inputs.

<!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `serving_funcs` </td> <td> A array of Jax functions with all the weights applied already. </td> </tr><tr> <td> `inputs` </td> <td> A array of Jax input placeholders tuples list, e.g., jnp.zeros(INPUT_SHAPE). Each tuple list should correspond with the serving function. </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> TFLiteConverter object. </td> </tr> </table> <h3 id="from_concrete_functions"><code>from_concrete_functions</code></h3>

<a target="_blank" class="external" href="https://github.com/tensorflow/tensorflow/blob/v2.11.0/tensorflow/lite/python/lite.py#L1710-L1746">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>@classmethod</code> <code>from_concrete_functions( funcs, trackable_obj=None ) </code></pre>

Creates a TFLiteConverter object from ConcreteFunctions.

<!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `funcs` </td> <td> List of TensorFlow ConcreteFunctions. The list should not contain duplicate elements. Currently converter can only convert a single ConcreteFunction. Converting multiple functions is under development. </td> </tr><tr> <td> `trackable_obj` </td> <td> An `AutoTrackable` object (typically `tf.module`) associated with `funcs`. A reference to this object needs to be maintained so that Variables do not get garbage collected since functions have a weak reference to Variables. </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> TFLiteConverter object. </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Raises</th></tr> <tr class="alt"> <td colspan="2"> Invalid input type. </td> </tr> </table> <h3 id="from_keras_model"><code>from_keras_model</code></h3>

<a target="_blank" class="external" href="https://github.com/tensorflow/tensorflow/blob/v2.11.0/tensorflow/lite/python/lite.py#L1814-L1828">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>@classmethod</code> <code>from_keras_model( model ) </code></pre>

Creates a TFLiteConverter object from a Keras model.

<!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `model` </td> <td> tf.Keras.Model </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> TFLiteConverter object. </td> </tr> </table> <h3 id="from_saved_model"><code>from_saved_model</code></h3>

<a target="_blank" class="external" href="https://github.com/tensorflow/tensorflow/blob/v2.11.0/tensorflow/lite/python/lite.py#L1748-L1812">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>@classmethod</code> <code>from_saved_model( saved_model_dir, signature_keys=None, tags=None ) </code></pre>

Creates a TFLiteConverter object from a SavedModel directory.

<!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `saved_model_dir` </td> <td> SavedModel directory to convert. </td> </tr><tr> <td> `signature_keys` </td> <td> List of keys identifying SignatureDef containing inputs and outputs. Elements should not be duplicated. By default the `signatures` attribute of the MetaGraphdef is used. (default saved_model.signatures) </td> </tr><tr> <td> `tags` </td> <td> Set of tags identifying the MetaGraphDef within the SavedModel to analyze. All tags in the tag set must be present. (default {tf.saved_model.SERVING} or {'serve'}) </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> TFLiteConverter object. </td> </tr> </table> <!-- Tabular view --> <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Raises</th></tr> <tr class="alt"> <td colspan="2"> Invalid signature keys. </td> </tr> </table>