Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance) #1544
Job | Run time |
---|---|
2s | |
1s | |
3m 37s | |
29m 23s | |
26m 2s | |
29m 22s | |
29m 6s | |
29m 23s | |
2h 26m 56s |
Job | Run time |
---|---|
2s | |
1s | |
3m 37s | |
29m 23s | |
26m 2s | |
29m 22s | |
29m 6s | |
29m 23s | |
2h 26m 56s |