model.py

class mlui.classes.model.Model

Bases: object

Class representing a machine learning model.

This class provides methods for managing and interacting with a TensorFlow machine learning model.

reset_state()

Reset the model state.

This method resets the internal state of the model, including its configuration and assigned features.

_set_config()

Set the configuration attributes for the model.

update_state()

Update the internal state of the model, resetting the input and output features.

_shapes_to_list(shapes)

Convert shapes from a dictionary or a single tuple to a list.

Parameters:

shapes (dict of tuples, list of tuples or tuple) – Input or output shapes. Single shape is a tuple of (None, int).

Returns:

Converted shapes.

Return type:

list of tuples

_get_processed_shape(at)

Retrieve and process the shapes of input or output layers.

Parameters:

at ({'input', 'output'}) – Side to retrieve shapes for.

Returns:

Processed shapes for the specified side.

Return type:

dict of {str to int}

_get_processed_data(data, at)

Process the input or output data based on the specified side.

Parameters:

• data (DataFrame) – Input or output data.

• at ({'input', 'output'}) – Side to process data for.

Returns:

Processed data.

Return type:

dict of {str to NDArray}

set_optimizer(entity, params)

Set the optimizer for the model.

Parameters:

• entity (str) – Name of the optimizer type.

• params (OptimizerParams) – Parameters for the optimizer.

Raises:

SetError – If there is an issue setting the optimizer.

get_optimizer()

Get the name of the current optimizer.

Returns:

Name of the optimizer.

Return type:

str or None

set_loss(layer, entity)

Set the loss function for a specific output layer.

Parameters:

• layer (str) – Name of the output layer.

• entity (str) – Name of the loss function type.

get_loss(layer)

Get the loss function for a specific output layer.

Parameters:

layer (str) – Name of the output layer.

Returns:

Name of the loss function.

Return type:

str or None

set_metrics(layer, entities)

Set the metrics for a specific output layer.

Parameters:

• layer (str) – Name of the output layer.

• entities (list of str) – Names of the metrics.

get_metrics(layer)

Get the metrics for a specific output layer.

Parameters:

layer (str) – Name of the output layer.

Returns:

Names of the metrics.

Return type:

list of str

compile()

Compile the model.

Raises:

ModelError – If there is an issue compiling the model.

set_features(layer, columns, at)

Set the input or output features for a specific layer.

Parameters:

• layer (str) – Name of the layer.

• columns (list of str) – Names of the columns.

• at ({'input', 'output'}) – Side to set features for.

Raises:

SetError – If there is an issue setting the features.

get_features(layer, at)

Get the input or output features for a specific layer.

Parameters:

• layer (str) – Name of the layer.

• at ({'input', 'output'}) – Side to get features for.

Returns:

Names of the features.

Return type:

list of str

set_callback(entity, params)

Set the callback for the model.

Parameters:

• entity (str) – Name of the callback type.

• params (CallbackParams) – Parameters for the callback.

Raises:

SetError – If there is an issue setting the callback.

get_callback(entity)

Get the callback for the model.

Parameters:

entity (str) – Name of the callback type.

Returns:

Callback instance.

Return type:

Callback or None

delete_callback(entity)

Delete the callback from the model.

Parameters:

entity (str) – Name of the callback type.

fit(data, batch_size, num_epochs, val_split)

Fit the model to the provided data.

Parameters:

• data (DataFrame) – Input and output data.

• batch_size (int) – Batch size.

• num_epochs (int) – Number of epochs.

• val_split (float) – Validation split.

Raises:

ModelError – If there is an issue fitting the model.

_update_history(logs)

Update the training history with new logs.

Parameters:

logs (DataFrame) – New logs to be added to the history.

plot_history(y, points)

Plot the training history.

Parameters:

• y (list of str) – Names of the logs to plot.

• points (bool) – Whether to include points on the plot.

Returns:

Altair chart representing the training history.

Return type:

Chart

property name: str

Name of the model.

property inputs: list[str]

Names of the input layers.

property outputs: list[str]

Names of the output layers.

property input_shape: dict[str, int]

Shapes of the input layers.

property output_shape: dict[str, int]

Shapes of the output layers.

property input_configured: bool

True if all input layers are configured, False otherwise.

property output_configured: bool

True if all output layers are configured, False otherwise.

property built: bool

True if the model is built, False otherwise.

property compiled: bool

True if the model is compiled, False otherwise.

property history: DataFrame

Training history DataFrame.

property summary: None

Summary of the model.

property graph: bytes

Bytes representation of the model graph.

property as_bytes: bytes

Bytes representation of the saved model.

class mlui.classes.model.UploadedModel

Bases: Model

Class representing the uploaded model.

upload(buff)

Upload a model from the provided file.

Parameters:

buff (file-like object) – Byte buffer containing the model.

Raises:

UploadError – If there is an issue reading the model from the file. If there is an issue validating the shapes of the model.

evaluate(data, batch_size)

Evaluate the model on the provided data.

Parameters:

• data (DataFrame) – Input and output data.

• batch_size (int) – Batch size.

Raises:

ModelError – If there is an issue evaluating the model.

Returns:

Evaluation results as a DataFrame.

Return type:

DataFrame

predict(data, batch_size)

Make predictions using the model on the provided data.

Parameters:

• data (DataFrame) – Input data.

• batch_size (int) – Batch size.

Raises:

ModelError – If there is an issue making predictions.

Returns:

Predictions as DataFrames.

Return type:

list of DataFrame

class mlui.classes.model.CreatedModel

Bases: Model

Class representing the created model.

reset_state()

Reset the state of the created model.

set_name(name)

Set the name of the created model.

Parameters:

name (str) – Name of the model.

set_layer(entity, name, params, connection)

Set the layer for the created model.

Parameters:

• entity (str) – Name of the layer type.

• name (str) – Name of the layer.

• params (LayerParams) – Parameters for the layer.

• connection (Layer, list of Layer or None) – Layer(s) to connect.

Raises:

SetError – If there is an issue setting the layer.

delete_last_layer()

Delete the last layer from the created model.

Raises:

DeleteError – If there are no layers to remove.

set_outputs(outputs)

Set the output layers for the created model.

Parameters:

outputs (list of str) – Names of the layers.

Raises:

SetError – If there is an issue setting the output layers.

create()

Create the machine learning model.

Raises:

CreateError – If there is an issue creating the model.

property layers: dict[str, Layer]

Objects of the layers.