module aixplain.factories.model_factory
class ModelFactory
A static class for creating and exploring Model Objects.
Attributes:
backend_url(str): The URL for the backend.
classmethod asset_repo_login
asset_repo_login(api_key: Optional[str] = None) → Dict
Return login credentials for the image repository that corresponds with the given API_KEY.
Args:
api_key(Text, optional): Team API key. Defaults to None.
Returns:
Dict: Backend response
classmethod create_asset_repo
create_asset_repo(
name: str,
description: str,
function: str,
source_language: str,
input_modality: str,
output_modality: str,
documentation_url: Optional[str] = '',
api_key: Optional[str] = None
) → Dict
Creates an image repository for this model and registers it in the platform backend.
Args:
name(Text): Model namehosting_machine(Text): Hosting machine ID obtained via list_host_machinesalways_on(bool): Whether the model should always be onversion(Text): Model versiondescription(Text): Model descriptionfunction(Text): Model function name obtained via LIST_HOST_MACHINESis_async(bool): Whether the model is asynchronous or not (False in first release)source_language(Text): 2-character 639-1 code or 3-character 639-3 language code.api_key(Text, optional): Team API key. Defaults to None.
Returns:
Dict: Backend response
classmethod create_utility_model
create_utility_model(
name: str,
code: Union[str, Callable],
inputs: List[UtilityModelInput] = [],
description: Optional[str] = None,
output_examples: str = ''
) → UtilityModel
Create a utility model
Args:
name(Text): name of the modelcode(Union[Text, Callable]): code of the modeldescription(Text, optional): description of the modelinputs(List[UtilityModelInput], optional): inputs of the modeloutput_examples(Text, optional): output examples
Returns:
UtilityModel: created utility model
classmethod deploy_huggingface_model
deploy_huggingface_model(
name: str,
hf_repo_id: str,
revision: Optional[str] = '',
hf_token: Optional[str] = '',
api_key: Optional[str] = None
) → Dict
Onboards and deploys a Hugging Face large language model.
Args:
name(Text): The user's name for the model.hf_repo_id(Text): The Hugging Face repository ID for this model ({author}/{model name}).hf_token(Text, optional): Hugging Face access token. Defaults to None.api_key(Text, optional): Team API key. Defaults to None.
Returns:
Dict: Backend response
classmethod get
get(model_id: str, api_key: Optional[str] = None) → Model
Create a 'Model' object from model id
Args:
model_id(Text): Model ID of required model.api_key(Optional[Text], optional): Model API key. Defaults to None.
Returns:
Model: Created 'Model' object
classmethod get_huggingface_model_status
get_huggingface_model_status(model_id: str, api_key: Optional[str] = None)
Gets the on-boarding status of a Hugging Face model with ID MODEL_ID.
Args:
model_id(Text): The model's ID as returned by DEPLOY_HUGGINGFACE_MODELapi_key(Text, optional): Team API key. Defaults to None.
Returns:
Dict: Backend response
classmethod list
list(
function: Optional[Function] = None,
query: Optional[str] = '',
suppliers: Optional[Supplier, List[Supplier]] = None,
source_languages: Optional[Language, List[Language]] = None,
target_languages: Optional[Language, List[Language]] = None,
is_finetunable: Optional[bool] = None,
ownership: Optional[Tuple[OwnershipType, List[OwnershipType]]] = None,
sort_by: Optional[SortBy] = None,
sort_order: SortOrder = <SortOrder.ASCENDING: 1>,
page_number: int = 0,
page_size: int = 20
) → List[Model]
Gets the first k given models based on the provided task and language filters
Args:
function(Function): function filter.source_languages(Optional[Union[Language, List[Language]]], optional): language filter of input data. Defaults to None.target_languages(Optional[Union[Language, List[Language]]], optional): language filter of output data. Defaults to None.is_finetunable(Optional[bool], optional): can be finetuned or not. Defaults to None.ownership(Optional[Tuple[OwnershipType, List[OwnershipType]]], optional): Ownership filters (e.g. SUBSCRIBED, OWNER). Defaults to None.sort_by(Optional[SortBy], optional): sort the retrived models by a specific attribute,page_number(int, optional): page number. Defaults to 0.page_size(int, optional): page size. Defaults to 20.
Returns:
List[Model]: List of models based on given filters
classmethod list_functions
list_functions(
verbose: Optional[bool] = False,
api_key: Optional[str] = None
) → List[Dict]
Lists supported model functions on platform.
Args:
verbose(Boolean, optional): Set to True if a detailed response is desired; is otherwise False by default.api_key(Text, optional): Team API key. Defaults to None.
Returns:
List[Dict]: List of dictionaries containing information about each supported function.
classmethod list_gpus
list_gpus(api_key: Optional[str] = None) → List[List[str]]
List GPU names on which you can host your language model.
Args:
api_key(Text, optional): Team API key. Defaults to None.
Returns:
List[List[Text]]: List of all available GPUs and their prices.
classmethod list_host_machines
list_host_machines(api_key: Optional[str] = None) → List[Dict]
Lists available hosting machines for model.
Args:
api_key(Text, optional): Team API key. Defaults to None.
Returns:
List[Dict]: List of dictionaries containing information about each hosting machine.
classmethod onboard_model
onboard_model(
model_id: str,
image_tag: str,
image_hash: str,
host_machine: Optional[str] = '',
api_key: Optional[str] = None
) → Dict
Onboard a model after its image has been pushed to ECR.
Args:
model_id(Text): Model ID obtained from CREATE_ASSET_REPO.image_tag(Text): Image tag to be onboarded.image_hash(Text): Image digest.host_machine(Text, optional): Machine on which to host model.api_key(Text, optional): Team API key. Defaults to None.
Returns:
Dict: Backend response