Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance) #1563
Job | Run time |
---|---|
3s | |
1s | |
14m 37s | |
14m 39s | |
14m 39s | |
14m 39s | |
14m 39s | |
14m 39s | |
1h 27m 56s |
Job | Run time |
---|---|
3s | |
1s | |
14m 37s | |
14m 39s | |
14m 39s | |
14m 39s | |
14m 39s | |
14m 39s | |
1h 27m 56s |