Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance) #1552
Job | Run time |
---|---|
3s | |
4s | |
29m 29s | |
30m 50s | |
18m 23s | |
30m 15s | |
42m 1s | |
36m 55s | |
3h 8m 0s |
Job | Run time |
---|---|
3s | |
4s | |
29m 29s | |
30m 50s | |
18m 23s | |
30m 15s | |
42m 1s | |
36m 55s | |
3h 8m 0s |