In the generation function for text generation models there is a parameter max_time. It makes sure the generation doesn’t last longer than x seconds. It also says that it finishes current pass and then stops. But what are the referenced passes? Does the prompt go multiple times through the model?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Text generation max length | 1 | 3113 | October 15, 2023 | |
| Generation_max_length, generation_num_beams meaning in seq2seq | 0 | 430 | December 25, 2023 | |
| Stopping generation before max_new_tokens | 0 | 804 | June 1, 2023 | |
| Pipeline max_length | 2 | 3977 | February 23, 2024 | |
| Output token lengths of smaller models | 0 | 505 | October 30, 2023 |