[Core] Subclass ModelRunner to support cross-attention & encoder sequences (towards eventual encoder/decoder model support) #17597
Job | Run time |
---|---|
1m 39s | |
1m 30s | |
1m 22s | |
1m 1s | |
1m 5s | |
6m 37s |
Job | Run time |
---|---|
1m 39s | |
1m 30s | |
1m 22s | |
1m 1s | |
1m 5s | |
6m 37s |