- configs
  - training configuration
- datasets
  - replica.py
  - scannet.py
- models
  - base.py # base abc model for sfrustum and gpnerf. should have basic interface, e.g. `loss()` for computing loss for training, `render()`, ... 
  - sfrustum.py
  - gpnerf.py
- modules
  - attention.py
  - ...
- utils
  - some impl of utils, e.g. make grid for ray
- train
  - trainer.py # basic trainer for controling data loading, training loop, etc. should be general and can be used by sfrustum and gpnerf. (it accept the `model` param)
- evaluation
  - evaluator.py # evaluation logic. should be general and can be used by sfrustum and gpnerf. (it accept the `model` param)
- tests
  - all test files to use. 
- scripts
  - shell scripts. can be empty for now.
main.py # entry point, can be used for training or inference.