# \[Bart\] Bart model families' embedding shape?

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-bart-model-families-embedding-shape%2F39750%3E%5C
**Category:** Beginners\
**Created:** [May 13, 2023, 1:08pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-bart-model-families-embedding-shape%2F39750 "2023-05-13T13:08:17Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![com3dian](https://avatars.discourse-cdn.com/v4/letter/c/e99b99/32.png) [@com3dian](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fcom3dian)%5C
**Post date:** [May 13, 2023, 1:08pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbart-bart-model-families-embedding-shape%2F39750%2F1 "2023-05-13T13:08:17Z")

</div>

Hi I’m implementing a finetuned Bart model for summarization, therefore I’m making decisions between using the ‘facebook/bart-large’ or the ‘facebook/bart-large-cnn’. But when I look into the layers in both models, I found the shapes of their embedding layers are different, is this a special trick?

Code to repeat

```python
from transformers import BartTokenizer, BartModel, BartForConditionalGeneration
BARTmodel = BartModel.from_pretrained('facebook/bart-large')
CGmodel = BartForConditionalGeneration.from_pretrained('facebook/bart-large-cnn')
BARTmodel.shared
-----------
Embedding(50265, 1024, padding_idx=1)

```

```python
CGmodel.model.shared
-----------
Embedding(50264, 1024, padding_idx=1)

```