MaximilianSchreff commented on PR #1941: URL: https://github.com/apache/systemds/pull/1941#issuecomment-1793369825
About caching: It would be a possibility to cache the normalized weights since they are quite complex to compute (counting of every degree and spectral convolution). In the stress testing, this only showed an improvement of 10 seconds from 340 seconds. Also, when testing the caching feature against smaller inputs than the huge inputs of the stress test, it was always slower. With the added complexity and difficulty to handle cached weights and generally more outputs of the layer, I decided against caching. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: dev-unsubscr...@systemds.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
