Implementing the CNN notebook

[hamidm21]

create a new branch for development and do the following:

1. load the textual dataset
2. create a pipeline for running them through cryptoBERT model to get the embeddings
3. stack the daily embeddings to make the new daily tweets dataset containing one embedding for each day with X * 768 dimensions. also, add padding to pad the embeddings to the max length in the dataset.
4. write the parallel CNN model described in Zou et al. paper.
5. run the embeddings through the CNN model to get the classes

[rajabi78]

using average is better than max to convert token level embeddings to sentence level embeddings