Edit model card

T5v1.1-Base Fine-Tuned on SQuAD for Question Generation

Model in Action:

import torch
from transformers import T5Tokenizer, T5ForConditionalGeneration

trained_model_path = 'ZhangCheng/T5v1.1-Base-Fine-Tuned-for-Question-Generation'
trained_tokenizer_path = 'ZhangCheng/T5v1.1-Base-Fine-Tuned-for-Question-Generation'

class QuestionGeneration:

    def __init__(self):
        self.model = T5ForConditionalGeneration.from_pretrained(trained_model_path)
        self.tokenizer = T5Tokenizer.from_pretrained(trained_tokenizer_path)
        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
        self.model = self.model.to(self.device)
        self.model.eval()

    def generate(self, answer:str, context:str):
        input_text = '<answer> %s <context> %s ' % (answer, context)
        encoding = self.tokenizer.encode_plus(
            input_text,
            return_tensors='pt'
        )
        input_ids = encoding['input_ids'].to(self.device)
        attention_mask = encoding['attention_mask'].to(self.device)
        outputs = self.model.generate(
            input_ids = input_ids,
            attention_mask = attention_mask
        )
        question = self.tokenizer.decode(
            outputs[0],
            skip_special_tokens = True,
            clean_up_tokenization_spaces = True
        )
        return {'question': question, 'answer': answer}

if __name__ == "__main__":
    context = 'ZhangCheng fine-tuned T5v1.1 on SQuAD dataset for question generation.'
    answer = 'ZhangCheng'
    QG = QuestionGeneration()
    qa = QG.generate(answer, context)
    print(qa['question'])
    # Output: 
    # Who fine-tuned T5v1.1 on SQuAD?
Downloads last month
13
Safetensors
Model size
248M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train ZhangCheng/T5v1.1-Base-Fine-Tuned-for-Question-Generation