Baunsgaard commented on PR #1847: URL: https://github.com/apache/systemds/pull/1847#issuecomment-1606338504
> @Baunsgaard Initially I'd thought there was a problem with my benchmark for loading data with pandas, but it seems that the script runs correctly, and that the converter for pandas to frames cannot handle loading large datasets in a reasonable amount of time, am I mistaken, is this a known problem? Yes this is a known problem for strings that we want to optimize. Hence the first step is to make quality measurements to start the process. I do not know if we have problems with other data types. > Im pretty sure I could come up with an easy fix for it by extending Py4JConverterUtils to take some sort of array format directly from pandas or dumping the pandas columns to numpy to reuse part of the existing ones from the Matrices. That defnitely would break the scope of this issue though. Yup it would break it and also go into another task already there to do ... exactly what you are proposing. We are aware that the performance is bad for Pandas currently. There are multiple reasons. But again. measurements first then we can fix the performance after. -- 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
