The first section of this article was weird to me. > They decided to scramble the data randomly, then sat back and watched to see how the Transformers would re

The first section of this article was weird to me. > They decided to scramble the data randomly, then sat back and watched to see how the Transformers would react. What does this mean? During training or inference? Obviously this would cause issues in training. Normally for time series data, you’d encode the relative position of the data point in the matrix passed into the input layer, so the transformer has that positional information.