Student project made during the deep learning class at CentraleSupélec. The project is about developing a Twitter bot detection model compliant with real time analysis and single tweet data.
Please first install poetry package manager in order to install dependencies.
Then run the following command to create your python environments and install all the dependencies (python > 3.8.x required):
poetry install
This package can train 2 types of models, a LSTM-based and a Bert-based. To do training it is required to first
download the data from Google Drive
and put it in the data
directory. The models are also available in the Google Drive directory.
To train a LSTM please consider the following example commands:
python lstm.py \
--cuda \
--data ./data \
--save-dir ./models/lstm \
--log-interval 500 \
--epochs 15 \
--batch-size 32 \
--seq-len 64 \
--emsize 200 \
--nhid 200 \
--nlayers 2 \
--dropout 0.5 \
--lr 0.0001 \
--clip 5.
To train a Bert model:
python bert.py \
--cuda \
--data ./data \
--save-dir ./models/bert \
--log-interval 200 \
--epochs 2 \
--batch-size 32 \
--seq-len 64 \
--lr 0.00001 \
--eps 0.00000001