problem history
[HF transformers] ValueError 'Asking to pad but the tokenizer does not have a padding token' when batch-tokenizing with decoder-only (Llama/GPT) tokenizers
Revision reasons are not recorded in this slice.