Open QvQKing opened 4 months ago
I also get this bug,wdo you have any solution?
Comment out line 313:with precision_scope("cuda"),half-precision results in non-release of gpu memory
Comment out line 313:with precision_scope("cuda"),half-precision results in non-release of gpu memory
Great, Thanks!
There is an issue with the code for batch testing, which continuously occupies graphics memory during runtime until it explodes. The problem is shown as follows: