You don't even need to deal with any XML formats or anything, they publish a complete dataset on Huggingface that's just a few lines to load in your Python trai

You don't even need to deal with any XML formats or anything, they publish a complete dataset on Huggingface that's just a few lines to load in your Python training script https://huggingface.co/datasets/wikimedia/wikipedia https://huggingface.co/datasets/wikimedia/wikipedia