Updating the loss aware reservoir training strategy.
Adding the option of latest_points, which are appended to each batch in the training.
The reservoir also only carries indices rather than the actual data, which makes it computationally more efficient.
In order to append data to single batches, the _train_epoch method was customized.
Updating the loss aware reservoir training strategy.
Adding the option of latest_points, which are appended to each batch in the training. The reservoir also only carries indices rather than the actual data, which makes it computationally more efficient. In order to append data to single batches, the _train_epoch method was customized.