Version: 0.93

snorkelflow.sdk.ExternalModel

class snorkelflow.sdk.ExternalModel(external_model_name, column_mappings, finetuning_provider_type=FinetuningProvider.AWS_SAGEMAKER)

Bases: object

init

__init__(external_model_name, column_mappings, finetuning_provider_type=FinetuningProvider.AWS_SAGEMAKER)

Class to represent a trained 3rd party model. Returns from ExternalModelTrainer.finetune

Name	Type	Default	Info
external_model_name	`str`		The name generated by the external model finetuning job.
column_mappings	`Dict[FineTuningColumnType, str]`		The column mappings from FineTuningColumnType to the specific column names in the dataset.
finetuning_provider_type	`FinetuningProvider`	`<FinetuningProvider.AWS_SAGEMAKER: 'sagemaker'>`	The 3rd party training service. Defaults to FinetuningProvider.AWS_SAGEMAKER.

Methods

`__init__`(external_model_name, column_mappings)	Class to represent a trained 3rd party model.
`inference`(datasource_uids[, x_uids, ...])	3rd Party Inference using the finetuned model.

inference(datasource_uids, x_uids=None, generation_config=None, deployment_config=None, prompt_template=None, sync=True)

3rd Party Inference using the finetuned model.

Parameters Parameters
Returns Returns: if (sync=True) returns the model source uid, else returns the job id
Return type Return type: str

Name	Type	Default	Info
datasource_uids	`List[int]`		The datasource uids to use for inference.
x_uids	`Optional[List[str]]`	`None`	Optional x_uids for filtering the datasources.
generation_config	`Optional[Dict[str, Any]]`	`None`	Optional generation configuration, e.g.: `{ "max_new_tokens": 300, "top_k": 50, "top_p": 0.8, "do_sample": True, "temperature": 1, }`
deployment_config	`Optional[Dict[str, Any]]`	`None`	Optional deployment configuration e.g.: `{ "instance_type": "ml.g5.12xlarge", "instance_count": 1, }`
prompt_template	`Optional[str]`	`None`	Optional prompt template to LLM inference. Defaults to {instruction}
sync	`bool`	`True`	Whether to wait for the inference job to complete before returning.