Is there any restrictions on the length of the model output?

Open szl666 opened this issue 2 years ago • 1 comments

I ran the ./scripts/run_evaluation_with_lora.sh with my testing data. However, the string length of the model output seems to have a limit (Some strings in the testing data is very long), which lead to the low testing accuracy, how to solve this problem?

Apr 19 '23 04:04 szl666

Hi, You might want to train a model on your own data with a larger length or cut off the test sentences to a short one (may hurt the test accuracy).

Apr 19 '23 07:04 shizhediao

adjusting --max_new_tokens may also help.

May 15 '23 00:05 shizhediao

This issue has been marked as stale because it has not had recent activity. If you think this still needs to be addressed please feel free to reopen this issue. Thanks

May 15 '23 00:05 shizhediao

Where is the '--max_new_tokens'?

May 22 '23 12:05 Irrevers