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

[MoE Refactor] Oracle Select FP8+NVFP4 Kernels In Priority (#32414)

Browse Source
This commit is contained in:
Robert Shaw
2026-01-21 08:22:33 -05:00
committed by GitHub
parent e14467be43
commit 42135d6898
82 changed files with 2710 additions and 1563 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/kernels/moe/test_moe.py
View File

@@ -18,7 +18,7 @@ from transformers import MixtralConfig
from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
import vllm.model_executor.layers.fused_moe # noqa
from tests.kernels.moe.utils import fused_moe
from tests.kernels.moe.utils import fused_moe, make_dummy_moe_config
from tests.kernels.utils import opcheck, stack_and_dev, torch_experts, torch_moe
from vllm._aiter_ops import rocm_aiter_ops
from vllm.config import VllmConfig, set_current_vllm_config
@@ -332,7 +332,7 @@ def test_fused_moe(
#
quant_config = FUSED_MOE_UNQUANTIZED_CONFIG
m_fused_moe_fn = modular_triton_fused_moe(quant_config)
m_fused_moe_fn = modular_triton_fused_moe(make_dummy_moe_config(), quant_config)
def m_fused_moe(
a: torch.Tensor,
@@ -437,7 +437,7 @@ def test_naive_block_assignment_moe(
#
quant_config = FUSED_MOE_UNQUANTIZED_CONFIG
m_fused_moe_fn = modular_triton_fused_moe(quant_config)
m_fused_moe_fn = modular_triton_fused_moe(make_dummy_moe_config(), quant_config)
def m_fused_moe(
a: torch.Tensor,