Added torchrun compatibility for distributet training across multiple GPUs in a single node (single instance) #1568
Job | Run time |
---|---|
3s | |
2s | |
3m 37s | |
49m 50s | |
28m 47s | |
24m 2s | |
30m 11s | |
36m 45s | |
2h 53m 17s |
Job | Run time |
---|---|
3s | |
2s | |
3m 37s | |
49m 50s | |
28m 47s | |
24m 2s | |
30m 11s | |
36m 45s | |
2h 53m 17s |