# Multimodal transformer

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmultimodal-transformer%2F37482%3E%5C
**Category:** Models\
**Created:** [April 23, 2023, 11:31am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmultimodal-transformer%2F37482 "2023-04-23T11:31:06Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![DanRKan](https://avatars.discourse-cdn.com/v4/letter/d/e47774/32.png) [@DanRKan](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FDanRKan)%5C
**Post date:** [April 23, 2023, 11:31am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmultimodal-transformer%2F37482%2F1 "2023-04-23T11:31:06Z")

</div>

Hi! I’m currently having social media data with 4 modalities - image, text(sentences and context-free text like hashtags), categories, and time-series based(posting data per post and also username who posted it). I explored the Huggingface models for multimodal transformers and found that all the models used only 2 modalities (text-tex or image-text or speech-text or graph transformers).

Can I use an image-text multimodal transformer and fine-tune it for my dataset with 4 modalities where i/p=post’s information grouped by the user? Any tips on whether it’ll be good enough and how to do it?