Late interaction models dimensions

Hello,

first of all, nice work!

I have been trying to understand the shapes of the colbert models.
as far as I have seen the _**colbert-ir/colbertv2.0**_ has a dimension of **(n_of_tokens - 1) x 768**
the jina model: **_jinaai/jina-colbert-v2_** has a dimension of **(n_of_tokens - 1) x 1024**

compared to the fastembed library the dimensions for both models is **n_of_tokes x 128**

the [example in the provided notebook](https://colab.research.google.com/drive/14FqLc0N_z92_VgL_zygWV5pJZkaskyk7?usp=sharing) is about colipali and there to the embedding dimension is 128.
Is there a compression step missing? or am I missing something else?

thanks!

Start with the provided ColPali notebook and trace how the ColBERT models are loaded and their embeddings are shaped. Compare those results with the dimensions reported by fastembed, then document whether a compression or projection step exists and what the expected dimensions are.

Written by the indexing model from the issue text.