Add Benchmark for DeepGEMM Group GEMM (#3993)
This commit is contained in:
@@ -211,6 +211,7 @@ def get_benchmark(tp_size):
|
||||
x = torch.randn((m, k), device="cuda", dtype=torch.bfloat16)
|
||||
y = torch.randn((n, k), device="cuda", dtype=torch.bfloat16)
|
||||
|
||||
# Preprocess data before benchmarking
|
||||
x_fp8, x_scale = per_token_cast_to_fp8(x)
|
||||
y_fp8, y_scale = per_block_cast_to_fp8(y)
|
||||
x_scale_col_major = get_col_major_tma_aligned_tensor(x_scale.clone())
|
||||
|
||||
Reference in New Issue
Block a user