Developer Portal

Appearance

module: sdk.training.training.py ​

Link to GitHub

Entities:

  1. Training

Class Training ​

extends: -

is extended by: -

description ​

Training is a class that is responsible for the training functionality of the service and interacting with it

attributes ​

  • training_daemon (ModuleType): The module containing the training daemon stubs.
  • training_daemon_grpc (ModuleType): The module containing the training daemon gRPC stubs.
  • training (ModuleType): The module containing the training stubs.
  • service_client (ServiceClient): The ServiceClient instance.
  • is_enabled (bool): Whether the training is enabled.
  • payment_strategy (TrainingPaymentStrategy): The payment strategy used for training.

methods ​

__init__ ​

Initializes the Training object. Imports the necessary modules and initializes the TrainingPaymentStrategy object. Sets the is_enabled attribute using the _check_training method.

args: ​
  • service_client (ServiceClient): The ServiceClient instance.
  • training_added (bool): Whether the training is added in service proto file. Defaults to False.
returns: ​
  • None

get_model_id_object ​

Converts the model ID from a string to an object from stub.

args: ​
  • model_id (str): The model ID to convert.
returns: ​
  • The stub object for the model ID. (Any)

create_model ​

Creates a necessary request object NewModelRequest(AuthorizationDetails, NewModel) from stub and calls the create_model grpc method using the _call_method method.

args: ​
  • method_name (str): The name of the service method for which we want to create a new model.
  • model_name (str): The name of the model.
  • model_description (str): The description of the model. Defaults to empty string.
  • is_public_accessible (bool): Whether the model is publicly accessible. Defaults to False.
  • addresses_with_access (list[str]): A list of addresses with access to the model. Defaults to empty list.
returns: ​
  • The newly created model. (Model)

validate_model_price ​

Creates a necessary request object AuthValidateRequest(AuthorizationDetails, model_id, training_data_link) from stub and calls the validate_model_price grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to validate.
returns: ​
  • Price of validating the model. (int)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

train_model_price ​

Creates a necessary request object CommonRequest(AuthorizationDetails, model_id) from stub and calls the train_model_price grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to train.
returns: ​
  • Price of training the model. (int)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

delete_model ​

Creates a necessary request object CommonRequest(AuthorizationDetails, model_id) from stub and calls the delete_model grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to delete.
returns: ​
  • Status of the model. (ModelStatus)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

get_training_metadata ​

Calls the get_training_metadata grpc method using the _call_method method with empty request.

returns: ​
  • Information about the training on the service. (TrainingMetadata)

get_all_models ​

Creates a necessary request object AllModelsRequest from stub and calls the get_all_models grpc method using the _call_method method. Use arguments as the filters to get all models.

args: ​
  • statuses (list[ModelStatus]): Statuses by which models need to be filtered. Defaults to None (no filter).
  • is_public (bool): Whether the models are public or not. Defaults to None (no filter).
  • grpc_method_name (str): The name of the gRPC method to call. Defaults to empty string (no filter).
  • grpc_service_name (str): The name of the gRPC service to call. Defaults to empty string (no filter).
  • model_name (str): The name of the model. Defaults to empty string (no filter).
  • created_by_address (str): The address of the user who created the model. Defaults to empty string (no filter).
returns: ​
  • List of models. (list[Model])

get_model ​

Creates a necessary request object CommonRequest(AuthorizationDetails, model_id) from stub and calls the get_model grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to get.
returns: ​
  • Price of training the model. (int)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

get_method_metadata ​

Creates a necessary request object MethodMetadataRequest(grpc_method_name, grpc_service_name, model_id) from stub and calls the get_method_metadata grpc method using the _call_method method. You can get metadata by method_name or model_id.

args: ​
  • method_name (str): The name of the service method for which we want to get metadata.
  • model_id (str): The model ID for which we want to get metadata.
returns: ​
  • Object with dateset requirements. (MethodMetadata)

update_model ​

Creates a necessary request object UpdateModelRequest from stub and calls the get_model grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to update.
  • model_name (str): New name of the model. (Optional)
  • description (str): New description of the model. (Optional)
  • addresses_with_access (list[str]): Updated list of addresses with access to the model. (Optional)
returns: ​
  • Updated model. (Model)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

upload_and_validate ​

Checks dataset using the _check_dataset method. Sets price and model id into payment strategy. Creates a generator that returns request objects UploadInput with file data one byte at a time. Calls the upload_and_validate grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to validate.
  • zip_path (str | Path | PurePath): The path to the dataset.
  • price (int): The price of the method call.
returns: ​
  • Status of the model. (ModelStatus)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

train_model ​

Sets price and model id into payment strategy. Creates a necessary request object UpdateModelRequest from stub and calls the train_model grpc method using the _call_method method.

args: ​
  • model_id (str): The model ID to train.
  • price (int): The price of the method call.
returns: ​
  • Status of the model. (ModelStatus)
raises: ​
  • NoSuchModelException: If the model with the specified ID does not exist.
  • GRPCException: If the gRPC call fails.

_call_method ​

Calls the specified gRPC training method and returns the response.

args: ​
  • method_name (str): The name of the gRPC method to call.
  • request_data (Any): The request data to pass to the gRPC method.
  • paid (bool): Whether the method is paid or not. Defaults to False.
returns: ​
  • Response from the gRPC method. (Any)
raises: ​
  • GRPCException: If the gRPC call fails.

_get_training_stub ​

Creates a gRPC stub for the training with gRPC channel.

args: ​
  • paid (bool): Whether the method is paid or not. Defaults to False.
returns: ​
  • gRPC stub. (Any)

_get_auth_details ​

Creates a necessary request object AuthorizationDetails from stub with user data and signature.

args: ​

-method_msg (ModelMethodMessage): The message of the method.

returns: ​
  • Authorization stub object. (Any)

_check_method_name ​

Checks if the specified method name is valid using get_services_and_messages_info method of the service client.

args: ​
  • method_name (str): The name of the method.
returns: ​
  • Service name and method name. (tuple[str, str])
raises: ​
  • WrongMethodException: If the method name is invalid.

_check_training ​

Checks if training is enabled in the service using get_training_metadata method.

returns: ​
  • Whether training is enabled or not. (bool)

_check_dataset ​

Checks the dataset for compliance with the requirements (which is obtained via the get_method_metadata method).

args: ​
  • model_id (str): The model ID.
  • zip_path (str | Path | PurePath): The path to the dataset.
returns: ​
  • None
raises: ​
  • WrongDatasetException: If the dataset is not valid.

_get_grpc_channel ​

Creates a gRPC channel for paid methods.

args: ​
  • base_channel (grpc.Channel): The base gRPC channel.
returns: ​
  • gRPC channel with interceptor for paid training methods. (grpc.Channel)