[CI] Prune back the number of tests in tests/kernels/* (#9932)

Signed-off-by: mgoin <michael@neuralmagic.com>
This commit is contained in:
Michael Goin
2024-11-05 16:02:32 -05:00
committed by GitHub
parent 02462465ea
commit 235366fe2e
10 changed files with 60 additions and 36 deletions

View File

@@ -11,6 +11,28 @@ from tests.kernels.utils import opcheck
from vllm import _custom_ops as ops
from vllm.platforms import current_platform
MNK_FACTORS = [
(1, 256, 128),
(1, 16384, 1024),
(1, 24576, 496),
(16, 256, 496),
(16, 16384, 128),
(16, 24576, 4096),
(32, 8192, 4096),
(32, 16384, 4096),
(33, 1024, 1024),
(33, 8192, 128),
(64, 2048, 496),
(64, 16384, 1024),
(100, 8192, 496),
(128, 32768, 4096),
(256, 4096, 4096),
(512, 256, 1024),
(512, 8192, 4096),
(512, 16384, 128),
(512, 24576, 128),
]
CUDA_DEVICES = [
f"cuda:{i}" for i in range(1 if torch.cuda.device_count() == 1 else 2)
]
@@ -116,9 +138,7 @@ def cutlass_int8_gemm_helper(m: int,
(out, a, b, scale_a, scale_b, bias))
@pytest.mark.parametrize("m", [1, 16, 32, 64, 128, 256, 512, 222, 100, 33])
@pytest.mark.parametrize("n", [2048, 4096, 8192, 16384, 24576, 256, 1024])
@pytest.mark.parametrize("k", [128, 496, 1024])
@pytest.mark.parametrize("m,n,k", MNK_FACTORS)
@pytest.mark.parametrize("per_act_token", [True, False])
@pytest.mark.parametrize("per_out_ch", [True, False])
@pytest.mark.parametrize("use_bias", [True, False])
@@ -129,9 +149,7 @@ 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", [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("m,n,k", MNK_FACTORS)
@pytest.mark.parametrize("per_act_token", [True, False])
@pytest.mark.parametrize("per_out_ch", [True, False])
@pytest.mark.parametrize("use_bias", [True, False])