I thought LLM were fed text only in their training data set? I’ve only done image classifiers and object detectors so I was assuming they must be trained with

I thought LLM were fed text only in their training data set? I’ve only done image classifiers and object detectors so I was assuming they must be trained with similar pure datasets.