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

[Feature] support sequence parallelism using compilation pass (#16155)

Browse Source 
Signed-off-by: cascade812 <cascade812@outlook.com>
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
This commit is contained in:
cascade
2025-04-27 06:29:35 -07:00
committed by GitHub
parent ed7a29d9f8
commit 690fe019f0
21 changed files with 1072 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
tests/compile/test_functionalization.py
View File

@@ -10,7 +10,7 @@ from vllm.compilation.fusion import (FUSED_OPS, FusionPass, QuantKey,
kFp8DynamicTokenSym, kFp8StaticTensorSym)
from vllm.compilation.fx_utils import find_auto_fn, find_auto_fn_maybe, is_func
from vllm.compilation.noop_elimination import NoOpEliminationPass
from vllm.config import CompilationConfig
from vllm.config import CompilationConfig, VllmConfig
from .backend import TestBackend
@@ -49,13 +49,15 @@ def test_fix_functionalization(model: str, quant_key: QuantKey,
do_fusion: bool):
torch.set_default_device("cuda")
config = CompilationConfig.PassConfig(enable_fusion=do_fusion,
enable_noop=True)
noop_pass = NoOpEliminationPass(config)
fusion_pass = FusionPass.instance(config)
vllm_config = VllmConfig()
vllm_config.compilation_config = CompilationConfig(pass_config= \
CompilationConfig.PassConfig(enable_fusion=do_fusion,
enable_noop=True))
noop_pass = NoOpEliminationPass(vllm_config)
fusion_pass = FusionPass.instance(vllm_config)
passes = [noop_pass, fusion_pass] if do_fusion else [noop_pass]
func_pass = FixFunctionalizationPass(config)
func_pass = FixFunctionalizationPass(vllm_config)
backend_func = TestBackend(*passes, func_pass)
backend_no_func = TestBackend(*passes)