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

[Frontend][torch.compile] CompilationConfig Overhaul (#20283): name change compilation level to compilation mode, deprecation compilation level (#26355)

Browse Source 
Signed-off-by: morrison-turnansky <mturnans@redhat.com>
Signed-off-by: Morrison Turnansky <mturnans@redhat.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
Morrison Turnansky
2025-10-14 22:51:16 -04:00
committed by GitHub
parent e66d787bce
commit 96b9aa5aa0
42 changed files with 270 additions and 248 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
tests/compile/test_config.py
View File

@@ -4,7 +4,7 @@ import pytest
from vllm.compilation.counter import compilation_counter
from vllm.config import CompilationConfig, CUDAGraphMode, VllmConfig
from vllm.config.compilation import CompilationLevel
from vllm.config.compilation import CompilationMode
from vllm.utils import _is_torch_equal_or_newer, is_torch_equal_or_newer
@@ -90,16 +90,16 @@ def test_use_cudagraphs(vllm_runner, monkeypatch, enabled):
# forked needed to workaround https://github.com/vllm-project/vllm/issues/21073
@pytest.mark.forked
def test_dynamo_as_is(vllm_runner, monkeypatch):
def test_stock_torch_compile(vllm_runner, monkeypatch):
# Disable multiprocessing so that the counter is in the same process
monkeypatch.setenv("VLLM_ENABLE_V1_MULTIPROCESSING", "0")
with (
compilation_counter.expect(dynamo_as_is_count=1),
compilation_counter.expect(stock_torch_compile_count=1),
# loading the model causes compilation (if enabled) to happen
vllm_runner(
"facebook/opt-125m",
compilation_config={"level": 1},
compilation_config={"mode": CompilationMode.STOCK_TORCH_COMPILE},
gpu_memory_utilization=0.4,
) as _,
):
@@ -112,11 +112,11 @@ def test_no_compilation(vllm_runner, monkeypatch):
# Disable multiprocessing so that the counter is in the same process
monkeypatch.setenv("VLLM_ENABLE_V1_MULTIPROCESSING", "0")
with (
compilation_counter.expect(num_graphs_seen=0, dynamo_as_is_count=0),
compilation_counter.expect(num_graphs_seen=0, stock_torch_compile_count=0),
# loading the model causes compilation (if enabled) to happen
vllm_runner(
"facebook/opt-125m",
compilation_config={"level": 0},
compilation_config={"mode": CompilationMode.NONE},
gpu_memory_utilization=0.4,
) as _,
):
@@ -130,7 +130,7 @@ def test_enforce_eager(vllm_runner, monkeypatch):
monkeypatch.setenv("VLLM_ENABLE_V1_MULTIPROCESSING", "0")
with (
compilation_counter.expect(num_graphs_seen=0, dynamo_as_is_count=0),
compilation_counter.expect(num_graphs_seen=0, stock_torch_compile_count=0),
# loading the model causes compilation (if enabled) to happen
vllm_runner(
"facebook/opt-125m", enforce_eager=True, gpu_memory_utilization=0.4
@@ -151,7 +151,7 @@ def test_splitting_ops_dynamic():
if is_torch_equal_or_newer("2.9.0.dev"):
config = VllmConfig(
compilation_config=CompilationConfig(
level=CompilationLevel.PIECEWISE,
level=CompilationMode.VLLM_COMPILE,
use_inductor_graph_partition=True,
splitting_ops=["vllm::unified_attention"],
)
@@ -163,7 +163,7 @@ def test_splitting_ops_dynamic():
# When attn_fusion pass enabled, splitting_ops now default to attention ops.
config = VllmConfig(
compilation_config=CompilationConfig(
level=CompilationLevel.PIECEWISE,
level=CompilationMode.VLLM_COMPILE,
pass_config={"enable_attn_fusion": True, "enable_noop": True},
custom_ops=["+quant_fp8"],
cudagraph_mode=CUDAGraphMode.PIECEWISE,
@@ -178,7 +178,7 @@ def test_splitting_ops_dynamic():
if is_torch_equal_or_newer("2.9.0.dev"):
config = VllmConfig(
compilation_config=CompilationConfig(
level=CompilationLevel.PIECEWISE,
level=CompilationMode.VLLM_COMPILE,
use_inductor_graph_partition=True,
pass_config={"enable_attn_fusion": True, "enable_noop": True},
custom_ops=["+quant_fp8"],