Summary

• [ ] Inherit from base.model.BaseModel and implement forward function with ONNX compatibility in mind.

• [ ] Implement training class inherited from BaseTrainer to customize training logic in each epoch.

• [ ] Create config.json with the structure consistent with your model and training classes.

• [ ] Run scripts/train.py --debug to test onnx.export.

• [ ] Run scripts/train.py -c config.json to save logs and best models.

• [ ] Implement all inference hooks described in Inference.

• [ ] Create config.json for inference to specify path to trained model and testing dataset.

• [ ] Have fun !