Stanford CS 271 Project
- Change data to PyTorch Dataset
- [DONE] Fill in dataset.py MyDataset
- Visualize/Explore data
- viz.py
- Add model
- models/pnunet.py
- Change loss/optimization
- train.py
- Train/Tune - Try different models.
- Start tensorboard using the
checkpoints/
folder withtensorboard --logdir=checkpoints/
- Start and stop training using
python train.py --checkpoint=<checkpoint name>
. The code should automatically resume training at the previous epoch and continue logging to the previous tensorboard. - Run
python test.py --checkpoint=<checkpoint name>
to get final predictions.
- checkpoints/ (Only created once you run train.py)
- data/
- metrics/
- models/
- layers/
- ...
- visualizers/
- args.py (Modify default hyperparameters here)
- dataset.py (Create Dataset here)
- metric_tracker.py
- models.py (You may opt to keep all your models in one place instead)
- preprocess.py (Do any preprocessing steps you want before loading the data)
- test.py
- train.py
- util.py
- viz.py (Create more visualizations if necessary)