This PR implements the Uncertainty Sampling Strategy (closes #63 ) + additional features like the:
option to reset weights after every active learning run
functionality to increase epoch size after every active learning run
improve metric logging
❗️ There is still a bug with the calculations of our batches when we have multiple workers. This is a general bug and not based on this PR and hopefully will be fixed with @jonaskordt newest changes.
This PR implements the Uncertainty Sampling Strategy (closes #63 ) + additional features like the:
❗️ There is still a bug with the calculations of our batches when we have multiple workers. This is a general bug and not based on this PR and hopefully will be fixed with @jonaskordt newest changes.