Question Answering
Initialize a HappyQuestionAnswering() object to perform question answering.
This model answers a question given a body of that’s text relevant to the questions.
The outputted answer is always a text-span with the provided information.
Initialization Arguments:
- model_type (string): specify the model name in all caps, such as “ROBERTA” or “ALBERT”
- model_name(string): below is a URL that contains potential models. MODELS
- use_auth_token (string): Specify the authentication token to load private models.
- trust_remote_code (bool): Allow for custom Python files to be used from the model_name location.
We recommend using “HappyQuestionAnswering(“ALBERT”, “mfeb/albert-xxlarge-v2-squad2”)” for the best performance
Example 3.0:
from happytransformer import HappyQuestionAnswering
# --------------------------------------#
happy_qa_distilbert = HappyQuestionAnswering("DISTILBERT", "distilbert-base-cased-distilled-squad") # default
happy_qa_albert = HappyQuestionAnswering("ALBERT", "mfeb/albert-xxlarge-v2-squad2")
# good model when using with limited hardware
happy_qa_bert = HappyQuestionAnswering("BERT", "mrm8488/bert-tiny-5-finetuned-squadv2")
happy_qa_roberta = HappyQuestionAnswering("ROBERTA", "deepset/roberta-base-squad2")
happy_qa__private_roberta = HappyQuestionAnswering("ROBERTA", "user-repo/roberta-base-squad2", use_auth_token="123abc")