What happens when using the DPO Trainer and there’s a data point with sequence length greater than the model’s sequence length? Does it get cut off from the beginning until it fits? Does it throw an error?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Token indices sequence length is longer than the specified maximum sequence length for this model | 1 | 5680 | July 21, 2023 | |
| In transformer, if the text exceeds max_seq_length, how to deal with it | 0 | 399 | January 19, 2024 | |
| TypeError: DPODataCollator.__init__() got an unexpected keyword argument 'max_prompt_length' | 0 | 81 | October 28, 2024 | |
| Some unintended things happen in Seq2SeqTrainer example | 3 | 1600 | November 30, 2020 | |
| Predictions with pipeline fails to truncate test set | 0 | 192 | January 23, 2024 |