transformers
Fix how we compute the final non-padding token for ForSequenceClassification models
#35911
Merged

Loading