ValueError: The batch received was empty, your model won't be able to train on it. Double-check that your training dataset contains keys expected by the model: args,kwargs,label_ids,label

[thetraintomars]

I just had this error as well, and the troubleshooting guide here was helpful:

Debugging the training pipeline - Hugging Face LLM Course

My problem was that I passed the original dataset to my trainer, not the tokenized one. Of course, now I have a different error to debug…