# Summarization with mT5

**URL:** <https://discuss.huggingface.co/t/summarization-with-mt5/6199>\
**Category:** 🤗Transformers\
**Created:** [May 16, 2021, 10:03am UTC](https://discuss.huggingface.co/t/summarization-with-mt5/6199 "2021-05-16T10:03:28Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![demegire](https://avatars.discourse-cdn.com/v4/letter/d/96bed5/32.png) [@demegire](https://discuss.huggingface.co/u/demegire)\
**Post date:** [May 16, 2021, 10:03am UTC](https://discuss.huggingface.co/t/summarization-with-mt5/6199/1 "2021-05-16T10:03:28Z")

</div>

Hello,

I am trying to do summarization with mT5 and when I use the official summarization colab which uses seq2seq trainer, the model outputs trash. You can see my [GitHub issue here](https://github.com/huggingface/transformers/issues/11735). Do you have any ideas on how to proceed with summarization using mT5?

---

<div class="post-metadata">

**Author:** ![demegire](https://avatars.discourse-cdn.com/v4/letter/d/96bed5/32.png) [@demegire](https://discuss.huggingface.co/u/demegire)\
**Post date:** [May 16, 2021, 10:51am UTC](https://discuss.huggingface.co/t/summarization-with-mt5/6199/2 "2021-05-16T10:51:32Z")

</div>

disabling fp16 seems to solve the issue, but I would rather have fp16 because it doubles the training speed 🙂