# Custom DistilBertTokenizer training

**URL:** <https://discuss.huggingface.co/t/custom-distilberttokenizer-training/1967>\
**Category:** 🤗Transformers\
**Created:** [November 10, 2020, 5:51pm UTC](https://discuss.huggingface.co/t/custom-distilberttokenizer-training/1967 "2020-11-10T17:51:18Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Dubin](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/dubin/32/561_2.png) [@Dubin](https://discuss.huggingface.co/u/Dubin)\
**Post date:** [November 10, 2020, 5:51pm UTC](https://discuss.huggingface.co/t/custom-distilberttokenizer-training/1967/1 "2020-11-10T17:51:18Z")

</div>

Hi All  
I am trying to find how to build a custom tokenizer for DistilBert all the examples I saw just use the pre-trained tokenizer.

> **[DistilBERT — transformers 3.5.0 documentation](https://huggingface.co/transformers/model_doc/distilbert.html)**

  
Can someone point me to how to build my custom model?  
Thanks in advance. 🙂

---

<div class="post-metadata">

**Author:** ![thomwolf](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/thomwolf/32/2198_2.png) [@thomwolf](https://discuss.huggingface.co/u/thomwolf)\
**Post date:** [November 12, 2020, 1:03pm UTC](https://discuss.huggingface.co/t/custom-distilberttokenizer-training/1967/2 "2020-11-12T13:03:03Z")

</div>

Hi, this is probably where you can start if you want to build a fast tokenizer: [https://huggingface.co/docs/tokenizers/python/master/quicktour.html](https://huggingface.co/docs/tokenizers/python/master/quicktour.html)

cc @anthony and @Narsil

---

<div class="post-metadata">

**Author:** ![Dubin](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/dubin/32/561_2.png) [@Dubin](https://discuss.huggingface.co/u/Dubin)\
**Post date:** [November 13, 2020, 8:10am UTC](https://discuss.huggingface.co/t/custom-distilberttokenizer-training/1967/3 "2020-11-13T08:10:13Z")

</div>

Thanks for the answer!. I will try the example it is a bit different and I wanted to make sure I have something more similar to the DistilBert.  
Thanks!

---

<div class="post-metadata">

**Author:** ![Dubin](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/dubin/32/561_2.png) [@Dubin](https://discuss.huggingface.co/u/Dubin)\
**Post date:** [November 13, 2020, 11:48am UTC](https://discuss.huggingface.co/t/custom-distilberttokenizer-training/1967/4 "2020-11-13T11:48:41Z")

</div>

I still do not see how to build this model as distilBertTokenizer any ideas?