llm-foundry
llm-foundry copied to clipboard
Why does convert_dataset_json.py only support 'train' for the --split argument?"
From the trace you posted, it looks like this is an issue with your dataset and/or HuggingFace's datasets library—the error is being raised by datasets/arrow_reader.py. My hunch is that there's no val split for your dataset, though I could be mistaken.