Enable FP8 Per-Tensor Scales and Integrate Marlin/MoE Kernels Repo for ROCm #2062
Job | Run time |
---|---|
8m 38s | |
7m 48s | |
5m 27s | |
5m 38s | |
7m 55s | |
2m 34s | |
0s | |
10m 18s | |
0s | |
0s | |
48m 18s |
Job | Run time |
---|---|
8m 38s | |
7m 48s | |
5m 27s | |
5m 38s | |
7m 55s | |
2m 34s | |
0s | |
10m 18s | |
0s | |
0s | |
48m 18s |