Don't use TP when tensor_parallel_degree
is 1 (#3636)
#10068
Job | Run time |
---|---|
3m 50s | |
3m 30s | |
3m 22s | |
10m 42s |
tensor_parallel_degree
is 1 (#3636)
#10068
Job | Run time |
---|---|
3m 50s | |
3m 30s | |
3m 22s | |
10m 42s |