This porject use the following features of a image posted on Reddit to predict it's popularity

• Title Embeddings
• Image Embeddings
• Text embeddings of caption generated from image
• Visual Embeddings (combined embedding of both image and text)
• Post Metadata
• Author's Metadata

The whole pipeline is managed using Apache Airflow. This includes

• Scrapping data
• Filtering data
• Fetching images
• Generating image captions
• Creating text, visual & combined embeddings
• Feature merging
• Training & saving the model

The pipeline is scheduled to run daily and train the model using new scrapped data. Each model trained is also saved with all it's evaluation metric.

Airlfow

export AIRFLOW_HOME=~/airflow

AIRFLOW_VERSION=2.9.1

# Extract the version of Python you have installed. If you're currently using a Python version that is not supported by Airflow, you may want to set this manually.
# See above for supported versions.
PYTHON_VERSION="$(python -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")')"

CONSTRAINT_URL="https://raw.githubusercontent.com/apache/airflow/constraints-${AIRFLOW_VERSION}/constraints-${PYTHON_VERSION}.txt"
# For example this would install 2.9.1 with python 3.8: https://raw.githubusercontent.com/apache/airflow/constraints-2.9.1/constraints-3.8.txt

pip install "apache-airflow==${AIRFLOW_VERSION}" --constraint "${CONSTRAINT_URL}"

FlagEmbedding

git clone https://github.com/FlagOpen/FlagEmbedding.git
cd FlagEmbedding
pip install -e .
pip install torchvision timm einops ftfy

You are also required to download the model for the Visual Embedding in order to generate combined embeddings. Download the model of your choice from https://huggingface.co/BAAI/bge-visualized and specify the path to weights in src\t06.4_create_embeddings_combined.py:50

Other dependencies

pip install -r requirements.txt

1. Clone the project

git clone https://github.com/DistilledCode/mmrl.git
cd mmrl

2. Create praw.ini & save Reddit credentials

[bot1]
client_id=secret
client_secret=secret
username=secret
password=secret

[bot2]
client_id=secret
client_secret=secret
password=secret
username=secret

Using 2 accounts is recommended, one for scraping and one for fetching the images. Using one account for both might give 429 response.

3. Start the scrapper. You may start the scrapper at any directory of your choice but monitor_scrapper.sh, scrap_comments.py & praw.ini should be in a single directory.

./monitor_scrapper.sh

4. Export the PROJ_DIR variable

cd mmrl
export PROJ_DIR=$PWD

5. Copy praw.ini and smpp_pipeline.py to ~/airflow/dags

cp praw.ini smpp_pipeline.py ~/airflow/dags
cp smpp_pipeline.py ~/airflow/dags

6. Start the Airflow scheduler from the PROJ_DIR directory

cd mmrl
airflow scheduler &

7. Start the Airflow Webserver & enable the pipeline

airflow webserver -p 8080 &

This will start the server at http:https://localhost:8080