CoolFace
Apppublic

kikwaib/mt5-base-kenswquad-abstractive

sourceHugging Faceapache-2.0updated 8mo agoView on Hugging Face
0likes
App README

KenSwQuAD Abstractive QA

This Space provides an interactive demo for abstractive question answering in Kenyan Swahili using the fine-tuned kikwaib/mt5-base-kenswquad-abstractive model.

How to Use

  1. 1.Enter Context: Paste or type your Swahili passage in the "Context" text area.
  2. 2.Enter Question: Type your question about the context in the "Question" field.
  3. 3.Get Answer: Click "Get Answer" to generate an abstractive answer.

Model Details

  • Base Model: mT5-base (~580M parameters)
  • Training Data: KenSwQuAD dataset (Kenyan Swahili)
  • Task: Abstractive Question Answering (seq2seq)

Notes

  • Supports both GPU and CPU inference (GPU recommended for faster responses).
  • Maximum context length: 1024 tokens (longer inputs will be truncated).
  • Generated answers are limited to 128 tokens.

Privacy

User submissions (context, question, answer) are logged for model evaluation purposes. Logs are retained temporarily and reviewed by administrators only.


Built with ❤️ for Swahili NLP