biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Kernel] Tuned int8 kernels for Ada Lovelace (#6848)

Browse Source 
Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com>
This commit is contained in:
Varun Sundar Rabindranath
2024-07-29 22:24:58 -04:00
committed by GitHub
parent 61a97c32f6
commit af647fb8b3
4 changed files with 395 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/kernels/test_cutlass.py
View File

@@ -119,8 +119,8 @@ def test_cutlass_fp8_gemm(m: int, n: int, k: int, per_act_token: bool,
cutlass_fp8_gemm_helper(m, n, k, per_act_token, per_out_ch, use_bias)
@pytest.mark.parametrize("m", [512, 222, 33, 1])
@pytest.mark.parametrize("n", [2048, 256, 1024])
@pytest.mark.parametrize("m", [1, 16, 32, 64, 128, 256, 512, 222, 33, 1])
@pytest.mark.parametrize("n", [2048, 8192, 16384, 256, 1024])
@pytest.mark.parametrize("k", [128, 496, 1024])
@pytest.mark.parametrize("per_act_token", [True, False])
@pytest.mark.parametrize("per_out_ch", [True, False])