1 If the user’s dataset does not contain enough sequences of the target length, sequences shorter than the target should be padded. 2 The `attention_mask` should be part of the dataset, not an attribute of AutoRound.
1 If the user’s dataset does not contain enough sequences of the target length, sequences shorter than the target should be padded.
2 The
attention_maskshould be part of the dataset, not an attribute of AutoRound.