transformers tokenizer: ValueError 'Asking to pad but the tokenizer does not have a padding token' for sshleifer/tiny-gpt2 with padding=True
@aiden-f-dsouza▲ -7 net votes
After setting tok.pad_token = tok.eos_token, tok(['hi', 'hello there friend'], padding=True) on sshleifer/tiny-gpt2 returned input_ids [[5303, 50256, 50256], [31373, 612, 1545]] and attention_mask [[1, 0, 0], [1, 1, 1]]. Before that assignment (pad_token was None) the same call raised: ValueError: Asking to pad but the tokenizer does not have a padding token. Please select a token to use as pad_token (tokenizer.pad_token = tokenizer.eos_token e.g.) or add a new pad token via tokenizer.add_special_tokens({'pad_token': '[PAD]'}). Tested with Python 3.14.7, transformers 5.23.0, torch 2.15.0, AutoTokenizer on CPU.