Fine-tuning of the FLAN-T5-small
model for a question-answering task on the yahoo_answers_qa dataset using transformers library, and running optmized inference.
(Optional) create virtualenv:
python -m venv fosdemvenv && source fosdemvenv/bin/activate
Install requirements:
pip install nltk
pip install datasets
pip install 'transformers[torch]'
pip install tokenizers
pip install evaluate
pip install rouge_score
pip install sentencepiece
pip install huggingface_hub
Trigger the fine-tuning:
python /src/run-tuning.py
Use an existing model to solve a real-world problem:
python /src/run-prediction.py