t5 large subjqa grocery qg

Introduction

The T5-LARGE-SUBJQA-GROCERY-QG model is a fine-tuned version of the lmqg/t5-large-squad specifically designed for the task of question generation using the lmqg/qg_subjqa dataset in the grocery domain. It leverages the T5 model architecture to generate questions from given passages or contexts.

Architecture

• Base Model: lmqg/t5-large-squad
• Language: English
• Pipeline Tag: Text2text Generation
• Tags: Question Generation

Training

The model was fine-tuned with the following hyperparameters:

• Dataset Path: lmqg/qg_subjqa
• Dataset Name: Grocery
• Input Types: ['paragraph_answer']
• Output Types: ['question']
• Prefix Types: ['qg']
• Model: lmqg/t5-large-squad
• Max Length: 512
• Max Length Output: 32
• Epochs: 3
• Batch Size: 16
• Learning Rate: 5e-05
• Floating Point Precision: 16-bit not used (fp16: False)
• Random Seed: 1
• Gradient Accumulation Steps: 32
• Label Smoothing: 0.15

Metrics for evaluating the model include BLEU4, METEOR, ROUGE-L, BERTScore, and MoverScore.

Guide: Running Locally

Basic Steps

1. Install Dependencies: Ensure you have transformers and lmqg libraries installed.

   pip install transformers
   pip install lmqg
   

2. Initialize the Model: Use the provided classes to load and interact with the model.

   from lmqg import TransformersQG
   model = TransformersQG(language="en", model="lmqg/t5-large-subjqa-grocery-qg")
   

3. Generate Questions: Use the model to generate questions based on provided context.

   questions = model.generate_q(list_context="William Turner was an English painter...", list_answer="William Turner")
   

4. Alternative with Transformers Pipeline:

   from transformers import pipeline
   pipe = pipeline("text2text-generation", "lmqg/t5-large-subjqa-grocery-qg")
   output = pipe("generate question: <hl> Beyonce <hl> further expanded her acting career...")
   

Cloud GPUs

For more efficient processing and training, consider using cloud-based GPU services like AWS, Google Cloud, or Azure to run the model.

License

This model is licensed under the Creative Commons Attribution 4.0 International (cc-by-4.0). This means you are free to share and adapt the model as long as appropriate credit is given, and any changes are indicated.