Closed elisaparga19 closed 5 months ago
Thank you for your interest! I only concatenated attention weights for plotting examples of attention over input windows later in running.py evaluate method. Attention is removed from the real output in all anomaly detection / seizure identification calculations.
Hello!
I'm adapting your work for anomaly detection in ECGs :) I just have one question with respect to the forward method of the TSTransformerEncoder. Why in the end you concatenate the output with the attention weights?
Thanks in advance!
Elisa.