report

## The Best model i build is on huggingface repo,it is pre-trained and fined tuned using labelled dataset manually made by me.
.Though it can't be uploaded to github as >500mb.The 'model-best' uploded at github is not pre-trained on unsupervised datas thus result in not too good an accuracy.

## LLama 2 didn't work for me due to not enough resources, even on google colab. It might classify better than 
current model in huggingface repo.

##Also no-shot wasn't able to classify tokens.