CJONS-4

Paths

├── kobert
│   ├── utils
│   │   ├── __init__.py
│   │   ├── aws_s3_downloader.py
│   │   └── utils.py
│   │
│   ├── __init__.py
│   └── pytorch_kobert.py
│
├── data
│   ├── data_info.pkl
│   ├── test.pkl
│   ├── train.pkl
│   └── valid.pkl
│
├── dataset
│   ├── photos
│   │   ├── --0h6FMC0V8aMtKQylojEg.jpg
│   │   └── --....jpg
│   │
│   ├── photos.json
│   ├── yelp_academic_dataset_business.json
│   ├── yelp_academic_dataset_review.json
│   ├── yelp_academic_dataset_user.json
│   ├── yelp_dataset.tar
│   └── yelp_photos.tar
│
├── model_parameters
│   ├── ncf.pt
│   ├── ncf_lstm.pt
│   └── mmr.pt
│
├── bpe_tokenizer.py
├── data_utils.py
├── Dockerfile
├── models.py
├── settings.py
├── train.py
├── utils.py
├── requirements.txt
└── README.md

Description

We're providing guidelines for Multi-Modal Recommender Systems with Anomaly Detection (For short MMR-AD) that is proposed by CJons-4 Team based on the datasets available at Yelp.com. We implemented MMR-AD by using PyTorch, Scikit-learn, Pandas, etc.

data_utils.py: includes Dataset, DataLoader.

models.py: includes LSTM, NCF, ResNet.

settings.py: includes configuration for setting paths.

utils.py: includes utilization function.

Unzip tarfile

from settings import * 
import tarfile, glob 

def unzip_tarfile(path):
    with tarfile.open(path, 'r') as f:
        f.extractall('dataset')
        
paths = glob.glob(DATA_DIR + '/*.tar')

for p in paths:
    unzip_tarfile(p)

Guide

1. Clone this repository

git clone https://github.com/ceo21ckim/CJONS-4.git

cd CJONS-4

2. Build Dockerfile

docker build --tag [filename]:1.0

3. Execute/run docker container

docker run -itd --gpus all --name cjons -p 8888:8888 -v C:\[PATH]\:/workspace [filename]:1.0 /bin/bash

4. Use jupyter notebook

docker exec -it [filename] bash

jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root

5. Training

python3 train.py --wandb \\
                 --lr 1e-3 \\
                 --num_epochs 100 \\
                 --batch_size 512\\
                 --hidden_dim 65 \\
                 --bidirectional \\
                 --dr_rate 0.2 \\
                 --max_len 128 \\
                 --size 256 \\
                 --model mmr \\
                 --device cuda:0 \\
                 --patience 3 

Training