matrice.models module#

class matrice.models.Model(session, model_id=None, model_name='')[source]#

Bases: object

The Model class provides methods for interacting with models in a project, including fetching summaries, listing models, and performing evaluations.

Parameters:

session (Session) – A session object containing the project ID and RPC client.
model_id (str, optional) – The unique identifier for the model (default is None).
model_name (str, optional) – The name of the model (default is an empty string).

Example

>>> session = Session(project_id="project123")
>>> model = Model(session, model_id="model789")

__init__(session, model_id=None, model_name='')[source]#

add_evaluation(dataset_id, dataset_version, split_types, is_pruned=False, is_gpu_required=False)[source]#

Add a new model evaluation using specified parameters.

Parameters:

dataset_id (str) – The ID of the dataset.
dataset_version (str) – The version of the dataset.
split_types (list) – The split types used in the evaluation.
is_pruned (bool, optional) – Whether the model is pruned (default is False).
is_gpu_required (bool, optional) – Whether the model requires a GPU (default is False).

Returns:

A tuple with the evaluation result, error message, and status message.

Return type:

tuple

Example

>>> result, error, message = model.add_model_eval(
>>>     id_dataset="dataset123",
>>>     dataset_version="v1.0",
>>>     split_types=["train", "val"],
>>> )

delete()[source]#

Delete the trained model.

Returns:: A tuple with the deletion result, error message, and status message.
Return type:: tuple

Example

>>> result, error, message = model.delete()
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Model deleted: {result}")

download_model(file_name)[source]#

Download the specified model type to a local file. There are 2 types of model types: trained and exported.

Parameters:

file_name (str) – The name of the file to save the downloaded model.
model_type (str) – The type of the model to download. Default is “trained”.

Returns:

A tuple with the download status, error message, and status message.

Return type:

tuple

Example

>>> result, error, message = model.download_model("model.pth", model_type="trained")
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Model downloaded: {result}")

get_details()[source]#

Get model details based on the provided ID or name.

Returns:: A tuple containing the model details, error message, and status message.
Return type:: tuple

Example

>>> details, error, message = model.get_details()
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Model details: {details}")

get_download_path()[source]#

Get the download path for the specified model type. There are 2 types of model types: trained and exported.

Parameters:: model_type (str) – The type of the model to download.
Returns:: A tuple with the download path, error message, and status message.
Return type:: tuple

Example

>>> download_path, error, message = model.get_model_download_path("trained")
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Download path: {download_path}")

get_eval_result(dataset_id, dataset_version, split_type)[source]#

Fetch the evaluation result of a trained model using a specific dataset version and split type.

Parameters:

dataset_id (str) – The ID of the dataset.
dataset_version (str) – The version of the dataset.
split_type (str) – The type of split used for the evaluation.

Returns:

A tuple with the evaluation result, error message, and status message.

Return type:

tuple

Example

>>> eval_result, error, message = model.get_eval_result("dataset123", "v1.0", "train")
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Evaluation result: {eval_result}")

get_model_training_logs()[source]#

Fetch training logs for the specified model.

This method retrieves the logs of the training epochs for a model, including both training and validation metrics such as losses and accuracy.

Returns:: A tuple containing: - A dictionary with the response from the RPC call. - An error message if the request fails. - A success message if the request succeeds.
Return type:: tuple

Example

>>> response, error, message = model_logging.get_model_training_logs()
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Success: {message}")

get_prediction(image_path)[source]#

Tests a trained model for a given image.

Parameters:#

image_pathstr: The path to the image for testing.

Returns:#

: tuple:

A tuple consisting of (result, error, message) with the test results.

Example:#

>>> result, error, message = model.test_model("/path/to/test_image.jpg")
>>> print(result)
{'test_result': 'success', 'confidence': 0.85}

model_test(model_type='trained')[source]#

Fetch information about the deployment server for a specific model.

Parameters:

model_train_id (str) – The ID of the model training instance.
model_type (str) – The type of model (e.g., ‘trained’, ‘exported’).

Returns:

A tuple containing three elements: - API response (dict): The raw response from the API. - error_message (str or None): Error message if an error occurred, None otherwise. - status_message (str): A status message indicating success or failure.

Return type:

tuple

Examples

>>> resp, err, msg = model.model_test("trained")
>>> if err:
>>>     print(f"Error: {err}")
>>> else:
>>>     print(f"Deployment server details : {resp}")

plot_epochs_losses()[source]#

Plot training and validation losses over epochs.

This method generates two subplots: one for the training losses and one for the validation losses, displaying how these metrics evolve over the epochs.

Return type:: None

Example

>>> model_logging.plot_epochs_losses()

plot_epochs_metrics()[source]#

Plot training and validation metrics (excluding losses) over epochs.

This method generates subplots for each non-loss metric, such as accuracy, showing how these metrics change during training epochs for both training and validation splits.

Return type:: None

Example

>>> model_logging.plot_epochs_metrics()

plot_eval_results()[source]#

Plot the evaluation results for the model.

Example

>>> model.plot_eval_results()

refresh()[source]#: Refresh the instance by reinstantiating it with the previous values.

rename(name)[source]#

Update the name of the trained model.

Parameters:: name (str) – The new name for the trained model.
Returns:: A tuple with the update result, error message, and status message.
Return type:: tuple

Example

>>> result, error, message = model.rename("NewModelName")
>>> if error:
>>>     print(f"Error: {error}")
>>> else:
>>>     print(f"Model name updated: {result}")