The readme does a great job for setting up / running a simple training loop. Some things are still missing, for example how to implement a new dataset. Perhaps it would be good to write the docs with several 'use cases' in mind:
- Inference on an existing model.
- Further training an existing model.
- Implementing a new dataset/use case.
- ..?
The readme does a great job for setting up / running a simple training loop. Some things are still missing, for example how to implement a new dataset. Perhaps it would be good to write the docs with several 'use cases' in mind: