Transformer Trainer example

Notebook example showing how to fine-tune Transformer models using the Transformer Trainer class.

Usage

Install the conda environment with

conda env create -f environment.yml

then log into the environment

conda activate transformers-trainer

and run Jupyter notebooks with

CUDA_VISIBLE_DEVICES=X jupyter notebook

where X is the number of the GPU you want to use for training.

Index of contents

• environment.yml: anaconda environment file
• transformers-multiclass.ipynb: notebook showint the usage of Transformer Trainer for a multiclass problem
• simplify_labels.ipynb: notebook used to simplify the dataset from multilabel to multiclass
• data: folder with multilabel and multiclass versions of the dataset