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

[FrontEnd] UNREVERT CompilationConfig overhaul (#20283): deprecate use_inductor in favor of backend, simplify custom_ops (#26502)

Browse Source 
Signed-off-by: morrison-turnansky <mturnans@redhat.com>
Signed-off-by: Morrison Turnansky <mturnans@redhat.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Jiangyun Zhu <riverclouds.zhu@qq.com>
This commit is contained in:
Morrison Turnansky
2025-10-13 18:47:16 -04:00
committed by GitHub
parent 7200a21cd1
commit e3fdb627d9
8 changed files with 153 additions and 86 deletions
Download Patch File Download Diff File Expand all files Collapse all files

39
tests/model_executor/test_enabled_custom_ops.py
View File

@@ -36,55 +36,56 @@ class Relu3(ReLUSquaredActivation):
@pytest.mark.parametrize(
"env, torch_level, use_inductor, ops_enabled, default_on",
"env, torch_level, backend, ops_enabled, default_on",
[
# Default values based on compile level
# - All by default (no Inductor compilation)
(None, 0, False, [True] * 4, True),
(None, 1, True, [True] * 4, True),
(None, 2, False, [True] * 4, True),
(None, 0, "eager", [True] * 4, True),
(None, 1, "eager", [True] * 4, True),
(None, 2, "eager", [True] * 4, True),
(None, 3, "eager", [True] * 4, True),
# - None by default (with Inductor)
(None, 3, True, [False] * 4, False),
(None, 4, True, [False] * 4, False),
# - All by default (without Inductor)
(None, 3, False, [True] * 4, True),
(None, 4, False, [True] * 4, True),
(None, 0, "inductor", [True] * 4, True),
# - None by default (with Inductor)
(None, 1, "inductor", [False] * 4, False),
(None, 2, "inductor", [False] * 4, False),
(None, 3, "inductor", [False] * 4, False),
# Explicitly enabling/disabling
#
# Default: all
#
# All but SiluAndMul
("+rms_norm,-silu_and_mul", 0, True, [1, 0, 1, 1], True),
("+rms_norm,-silu_and_mul", 0, "inductor", [1, 0, 1, 1], True),
# Only ReLU3
("none,-rms_norm,+relu3", 1, False, [0, 0, 0, 1], False),
("none,-rms_norm,+relu3", 1, "eager", [0, 0, 0, 1], False),
# All but SiluAndMul
("all,-silu_and_mul", 2, True, [1, 0, 1, 1], True),
("all,-silu_and_mul", 2, "inductor", [1, 0, 1, 1], True),
# All but ReLU3 (even if ReLU2 is on)
("-relu3,+relu2", 3, False, [1, 1, 1, 0], True),
("-relu3,+relu2", 3, "eager", [1, 1, 1, 0], True),
# RMSNorm and SiluAndMul
("none,-relu3,+rms_norm,+silu_and_mul", 4, False, [1, 1, 0, 0], False),
("none,-relu3,+rms_norm,+silu_and_mul", 3, "eager", [1, 1, 0, 0], False),
# All but RMSNorm
("-rms_norm", 3, False, [0, 1, 1, 1], True),
("-rms_norm", 3, "eager", [0, 1, 1, 1], True),
#
# Default: none
#
# Only ReLU3
("-silu_and_mul,+relu3", 3, True, [0, 0, 0, 1], False),
("none,+relu3", 3, "inductor", [0, 0, 0, 1], False),
# All but RMSNorm
("all,-rms_norm", 4, True, [0, 1, 1, 1], True),
("all,-rms_norm", 3, "inductor", [0, 1, 1, 1], True),
],
)
def test_enabled_ops(
env: str | None,
torch_level: int,
use_inductor: bool,
backend: str,
ops_enabled: list[int],
default_on: bool,
):
custom_ops = env.split(",") if env else []
vllm_config = VllmConfig(
compilation_config=CompilationConfig(
use_inductor=bool(use_inductor), level=torch_level, custom_ops=custom_ops
backend=backend, level=torch_level, custom_ops=custom_ops
)
)
with set_current_vllm_config(vllm_config):