# Trianing a model using predefined vocab

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ftrianing-a-model-using-predefined-vocab%2F21697%3E%5C
**Category:** Beginners\
**Created:** [August 16, 2022, 4:37pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ftrianing-a-model-using-predefined-vocab%2F21697 "2022-08-16T16:37:11Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![ockapuh](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/ockapuh/32/33342_2.png) [@ockapuh](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fockapuh)%5C
**Post date:** [August 16, 2022, 4:37pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ftrianing-a-model-using-predefined-vocab%2F21697%2F1 "2022-08-16T16:37:11Z")

</div>

Hi everyone.

I would like to train from scratch a model that uses as vocabulary a list of predefined tokens that I have. This way the vocab and te tokenizer wouldn’t be created during language modelling, and this training would only require to generate the predefined tokens embeddings. I am not sure of it it is possible, since I suppose if would require some kind of personalized tokenizer.

Thanks.