Spaces:
Running
Running
microsoft/harrier-oss-v1-27b evaluation
#2
by nie3e - opened
Unfortunately, we will not be including embedding models of that size. PIRB consists of 41 tasks, and some datasets contain millions of documents, so evaluating a 27B model would require too much compute and time. Around 10B is the maximum we are able to handle.
I agree with that - large data set will be a problem
nie3e changed discussion status to closed