Is there the code for downstream task in this project?
Claymore715 opened this issue · comments
to compare the efficiency between manual prompt and autoprompt
for example,how to test the performance for task Sentiment Analysis
For evaluation of prompts for fact retrieval and relation extraction, refer to this section of the README. For the other tasks, you'll have to write some code that will load the eval dataset (SST-2 or SICK-E), append the trigger tokens, feed the full prompt into the LM, and score the results.