Unexpected evaluation results with zeroed Loss (anomaly detection) #40
Description
Hello, I have been studying the ts2vec model recently. When I use the yahoo dataset for training, I encountered an inconsistency about the contribution of the loss. Even if i set the loss manually to zero during the training phase, the evaluation results are almost the same, sometimes even better. I was wondering if it is something that I am missing or if it is a possible issue encountered, perhaps related to the evaluation phase itself. I appreciate any guidance or assistance you can provide regarding this matter and I hope to receive your answer. Thank you for your time and attention to this issue!