README.md

Tabular approach

Structure

│
├── configs                   <- Hydra configs
│   ├── model                    <- Model configs
│   ├── paths                    <- Project paths configs
│   ├── wandb                    <- W&B configs
│   ├── sweep                    <- Sweep configs
│   │
│   └──config.yaml               <- Main config
│
├── tabular                    <- Source code
│   ├── feature                     <- Feature Engineering scripts
│   ├── data.py                     <- Data scripts
│   ├── metric.py                   <- Metric scripts
│   ├── trainer.py                  <- Trainer scripts
│   └── utils.py                    <- Utility scripts
│
├── main.py                  <- Run training and inference
├── test_main.py             <- Run testing with dummy data
├── requirements.txt         <- File for installing python dependencies
└── README.md

Setup

cd ~/level2_dkt_recsys-03/tabular/
conda init
(base) . ~/.bashrc
(base) conda create -n tabular python=3.10 -y
(base) conda activate tabular
(tabular) pip install -r requirements.txt

실행 명령어

python main.py

Sweep

wandb sweep {configs/sweep/sweep.yaml}
wandb agent {entity}/{project}/{sweepid}

# sweep.yaml example
project: DKT
entity: recsys01
program: main.py
method: grid
metric:
  goal: maximize
  name: cv_score
parameters:
  features:
    values: [default, comb1]

command:
  - ${env}
  - ${interpreter}
  - ${program}
  - ${args_no_hyphens}