Note that this repository is forked from github.com/ds4dm/ml4co-competition. See here for the official competition website.
For a related application, trying to explore the dataset using Bayesian Optimization, check out the bo_gnn_seminar branch.
- rank 3 / 15 on the global leaderboard
- rank 1 / 7 on the student leaderboard
Read our short write-up.
See also the pdf version.
.
├── bo_gnn <-- main directory
│ └── config
│ ├── conda.yaml
│ ├── config_utils.py
│ ├── data_utils
│ │ ├── dataset.py <-- dataset loader
│ │ └── milp_data.py <-- bipartite graph representation
│ ├── Dockerfile.cpu <-- Dockefile for data generation
│ ├── Dockerfile.gpu <-- Dockerfile for GNN training
│ ├── generate_data.py <-- massively parallel data generation script
│ ├── models
│ │ ├── baseline.py <-- GNN definition
│ │ └── callbacks.py
│ ├── mps_to_pkl.py
│ ├── plot_run_performance.py
│ ├── scripts
│ └── train_gnn.py <-- main training loop & ensembling
├── common <-- pre-defined evaluation environments
│ ├── environments.py
│ ├── evaluate.py
│ └── rewards.py
├── instances <-- raw data location
├── README.md
├── submissions <-- submission code
│ └── MixedInspiringLamePuns
│ ├── agents
│ ├── callbacks.py
│ ├── conda.yaml
│ ├── data_utils
│ ├── heuristics_schedules
│ ├── init.sh
│ ├── models
│ ├── param_configurations
│ ├── parameter_configuration_mapping
│ ├── results
│ ├── trained_model_checkpoints
│ └── train_gnn.py
└── TASKS.md