# BART for Portuguese

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%3E%5C
**Category:** 🤗Transformers\
**Created:** [July 23, 2020, 4:21pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448 "2020-07-23T16:21:54Z")\
**Posts on this page:** 8\
**Page:** 1

<div class="post-metadata">

**Author:** ![bunoviske](https://avatars.discourse-cdn.com/v4/letter/b/b2d939/32.png) [@bunoviske](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fbunoviske)%5C
**Post date:** [July 23, 2020, 4:21pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F1 "2020-07-23T16:21:54Z")

</div>

Hey @sshleifer !

I have a question: I need to perform Summarization on Portuguese texts. Is there some pre-trained BART for this language or some Multi-Lingual BART model?

If not, what is the recommended approach? I could think in some possibilities, like using portuguese encoder or translating the text to english…

Thanks!

---

<div class="post-metadata">

**Author:** ![sshleifer](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/sshleifer/32/3304_2.png) [@sshleifer](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fsshleifer)%5C
**Post date:** [July 23, 2020, 7:39pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F2 "2020-07-23T19:39:55Z")

</div>

There is no pretrained model for portuguese summarization, you would need to finetune multilingual bart (mbart). Do you have document/summary pairs to use as training data?

---

<div class="post-metadata">

**Author:** ![bunoviske](https://avatars.discourse-cdn.com/v4/letter/b/b2d939/32.png) [@bunoviske](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fbunoviske)%5C
**Post date:** [July 24, 2020, 8:00am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F3 "2020-07-24T08:00:07Z")

</div>

Ok! I still don’t have it, but I am preparing. Should I use BartForConditionalGeneration? If you could share one notebook example for finetunning this task would be great. Thanks.

---

<div class="post-metadata">

**Author:** ![sshleifer](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/sshleifer/32/3304_2.png) [@sshleifer](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fsshleifer)%5C
**Post date:** [July 27, 2020, 3:04pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F4 "2020-07-27T15:04:35Z")

</div>

I would use mbart, I will share a command when this [PR](https://github.com/huggingface/transformers/pull/6003) is merged.

---

<div class="post-metadata">

**Author:** ![thiagocmoreira](https://avatars.discourse-cdn.com/v4/letter/t/b4bc9f/32.png) [@thiagocmoreira](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fthiagocmoreira)%5C
**Post date:** [September 9, 2020, 7:28pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F5 "2020-09-09T19:28:03Z")

</div>

@bunoviske Were you able to train the model? If so, do you have a notebook to share? I need to do this summary with BART in Portuguese too, but I didn’t find anything about it.

---

<div class="post-metadata">

**Author:** ![BramVanroy](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/bramvanroy/32/47360_2.png) [@BramVanroy](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FBramVanroy)%5C
**Post date:** [September 9, 2020, 9:34pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F6 "2020-09-09T21:34:45Z")

</div>

@thiagocmoreira @bunoviske Rather than doing separate models, you may want to communicate with each other and share the training data and cost to fine tune mbart on Portuguese.

---

<div class="post-metadata">

**Author:** ![pierreguillou](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/pierreguillou/32/19626_2.png) [@pierreguillou](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fpierreguillou)%5C
**Post date:** [September 11, 2020, 6:23pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F7 "2020-09-11T18:23:42Z")

</div>

> [@BramVanroy](#):
>
> share the training data and cost to fine tune mbart on Portuguese

I agree with you @BramVanroy 🙂

Hello @bunoviske and @thiagocmoreira. There is a [list of datasets in Portuguese](https://forum.ailab.unb.br/t/datasets-em-portugues/251) in the [AI Lab forum](https://forum.ailab.unb.br/). When your dataset to train NLP models for Portuguese summary is ready, could you share its link in the corresponding topic of this forum as well? Thanks in advance!

---

<div class="post-metadata">

**Author:** ![bunoviske](https://avatars.discourse-cdn.com/v4/letter/b/b2d939/32.png) [@bunoviske](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fbunoviske)%5C
**Post date:** [October 20, 2020, 10:54am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-for-portuguese%2F448%2F8 "2020-10-20T10:54:00Z")

</div>

Hey! Sorry, currently I am not working anymore on this issue but it would be great if you update us about your current status and progress. Good luck!
