[GPU] Do not use usm_host when network output tensor size is large #11769
Job | Run time |
---|---|
21s | |
1m 8s | |
12m 9s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
2m 42s | |
1m 10s | |
6m 35s | |
4m 35s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
1s | |
28m 41s |
Job | Run time |
---|---|
21s | |
1m 8s | |
12m 9s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
2m 42s | |
1m 10s | |
6m 35s | |
4m 35s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
0s | |
1s | |
28m 41s |