I note in this https://github.com/Dao-AILab/quack/blob/main/benchmarks/benchmark_gemm_sm90.py, Since most fp8 Gemm case, we need to do it with A/B's scale(column major scale). If possible, maybe we can add scale to FP8 Gemm case?
I note in this https://github.com/Dao-AILab/quack/blob/main/benchmarks/benchmark_gemm_sm90.py,
Since most fp8 Gemm case, we need to do it with A/B's scale(column major scale). If possible, maybe we can add scale to FP8 Gemm case?