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

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
tests/kernels/moe/test_batched_deepgemm.py
View File

@@ -5,13 +5,14 @@ import pytest
import torch
from vllm.model_executor.layers.fused_moe.batched_deep_gemm_moe import (
BatchedDeepGemmExperts)
from vllm.model_executor.layers.fused_moe.config import (
fp8_w8a8_moe_quant_config)
BatchedDeepGemmExperts,
)
from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_config
from vllm.model_executor.layers.fused_moe.fused_batched_moe import (
BatchedPrepareAndFinalize, BatchedTritonExperts)
from vllm.model_executor.layers.fused_moe.modular_kernel import (
FusedMoEModularKernel)
BatchedPrepareAndFinalize,
BatchedTritonExperts,
)
from vllm.model_executor.layers.fused_moe.modular_kernel import FusedMoEModularKernel
from vllm.utils.deep_gemm import calc_diff, is_deep_gemm_supported
from .test_deepgemm import make_block_quant_fp8_weights
@@ -19,15 +20,15 @@ from .test_deepgemm import make_block_quant_fp8_weights
BLOCK_SIZE = [128, 128]
@pytest.mark.skipif(not is_deep_gemm_supported(),
reason="Requires deep_gemm kernels")
@pytest.mark.skipif(not is_deep_gemm_supported(), reason="Requires deep_gemm kernels")
@pytest.mark.parametrize("E", [16, 32]) # number of experts
@pytest.mark.parametrize("T", [256, 512]) # tokens per expert
@pytest.mark.parametrize("K", [128, 256]) # hidden dim
@pytest.mark.parametrize("N", [512, 1024]) # intermediate dim per expert
@pytest.mark.parametrize("topk", [2, 4])
def test_batched_deepgemm_vs_triton(E: int, T: int, K: int, N: int, topk: int,
monkeypatch):
def test_batched_deepgemm_vs_triton(
E: int, T: int, K: int, N: int, topk: int, monkeypatch
):
"""Compare BatchedDeepGemmExperts to BatchedTritonExperts."""
monkeypatch.setenv("VLLM_USE_DEEP_GEMM", "1")