graph-ml

Machine Learning On Graphs

.
├── exec                            # Executable files
│   ├── greeting.py 
├── architectures                   # Architecture-related classes 
│   ├── __init__.py   
│   ├── abstract_architecture_builder.py    # Interface for a builder 
│   ├── architecture_builder.py             # Basic builder example 
│   ├── architecture.py                     # Class to contain components 
│   ├── director.py                         # Class controlling the builder 
├── data                            # Data-related classes 
│   ├── __init__.py            
│   ├── abstract_data_factory.py            # Interface for a data factory 
│   ├── pyg_data_factory.py                 # Class for torch_geometric.datasets
│   ├── planetoid_data_factory.py           # Class for torch_geometric Planetoid 
├── experiments                     # Setups for ML tasks
│   ├── __init__.py
│   ├── abstract_experiment.py              # Interface for an experiment
│   ├── experiment.py                       # Basic experiment example
├── slurm                           # Slurm scripts & logs
│   ├── experiment.sh                       # Run experiment
│   ├── greeting.sh
├── tests                           # Tests for classes 
│   ├── director_test.py   
│   ├── experiment_test.py   
│   ├── planetoid_data_factory_test.py     
└── README.md

greeting.py is an hello world program that creates a Neptune run for the project, connects python logger to the run, logs cmd arguments and slurm output file names.

Working from remote server

Submit slurm scripts when pwd is the root directory.

Tests

Locate to root an run the tests as modules e.g. python -m tests.planetoid_data_factory_text

TODO

• Implement the generic class PyGDataFactory
• PlanetoidDataFactory should create DataLoader
• Experiment:construct_framework_architecture: Take spec file for architecture configuration