Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance) #1546
Job | Run time |
---|---|
2s | |
2s | |
29m 19s | |
31m 35s | |
38m 53s | |
3m 37s | |
25m 57s | |
36m 36s | |
2h 46m 1s |
Job | Run time |
---|---|
2s | |
2s | |
29m 19s | |
31m 35s | |
38m 53s | |
3m 37s | |
25m 57s | |
36m 36s | |
2h 46m 1s |