CPU/CUDA: fix (GQA) mul mat back, add CUDA support (#11380) #18
Job | Run time |
---|---|
12m 56s | |
3m 5s | |
6m 15s | |
20m 23s | |
12m 28s | |
5m 5s | |
18m 35s | |
2m 43s | |
3m 54s | |
12m 24s | |
6m 7s | |
1m 25s | |
1m 39s | |
3m 56s | |
2m 2s | |
5m 8s | |
1m 58s | |
11m 52s | |
43m 20s | |
2m 1s | |
39m 58s | |
4m 52s | |
6m 4s | |
5m 27s | |
5m 17s | |
5m 28s | |
6m 58s | |
4m 45s | |
5m 44s | |
6m 22s | |
3m 5s | |
3m 23s | |
3m 43s | |
8m 11s | |
1m 25s | |
26m 42s | |
24m 20s | |
30m 3s | |
11m 29s | |
32m 20s | |
2m 15s | |
6h 55m 7s |