Closed WEIYanbin1999 closed 1 year ago
Could you please provide a link for downloading the pre-trained model parameters? I think it is much helpful for us to follow your work
I had the same problem. I ran it on the cloud server RTX 3090 for 11 minutes before it was killed
The memory occupancy comes from the huge 2-hop triple data dict.
When I run the pretrain code, I found the memory will be toughly full in a machine with 100+G memory. Why its CPU occupancy and memory requirements are so high? Is it normal?
Even I running in another machine with 30G memory, it is killed by OS because of its memory payload