# T5/BART decoder prefix

**URL:** <https://discuss.huggingface.co/t/t5-bart-decoder-prefix/18909>\
**Category:** 🤗Transformers\
**Created:** [June 9, 2022, 11:28am UTC](https://discuss.huggingface.co/t/t5-bart-decoder-prefix/18909 "2022-06-09T11:28:31Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![sunhaozhepy](https://avatars.discourse-cdn.com/v4/letter/s/50afbb/32.png) [@sunhaozhepy](https://discuss.huggingface.co/u/sunhaozhepy)\
**Post date:** [June 9, 2022, 11:28am UTC](https://discuss.huggingface.co/t/t5-bart-decoder-prefix/18909/1 "2022-06-09T11:28:31Z")

</div>

I know that for models like GPT-2, it is possible to make a prefix, or a prompt, to let the model continue on this sequence of text. I wonder if we could do the same with encoder-decoder models, e.g., T5, BART…? (Note that this prefix is the decoder prefix, not the encoder prefix like in “Translate English to French: I’m hungry”)