[Bugfix] fix nightly multi-node EPLB tests' "DYNAMIC_EPLB=true" environment not working (#4223)
### What this PR does / why we need it?
fix nightly multi-node EPLB tests by adjusting
vllm_ascend\eplb\core\eplb_utils.py dynamic_eplb gate checking
### Does this PR introduce _any_ user-facing change?
no
- vLLM version: v0.11.0
- vLLM main:
2918c1b49c
---------
Signed-off-by: 白永斌 <baiyongbin3@h-partners.com>
Signed-off-by: 欧派果奶我还要 <47294568+845473182@users.noreply.github.com>
Co-authored-by: 白永斌 <baiyongbin3@h-partners.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
@@ -128,7 +128,7 @@ class TestEPLBParamUtils:
|
||||
with pytest.raises(
|
||||
ValueError,
|
||||
match=
|
||||
'Can not enable dynamic_eplb when not export DYNAMIC_EPLB="true".'
|
||||
'Can not enable dynamic_eplb when DYNAMIC_EPLB is not set to "true" or "1".'
|
||||
):
|
||||
EPLBParamUtils.check_dynamic_eplb(True)
|
||||
|
||||
@@ -136,7 +136,7 @@ class TestEPLBParamUtils:
|
||||
with pytest.raises(
|
||||
ValueError,
|
||||
match=
|
||||
'Can not enable dynamic_eplb when not export DYNAMIC_EPLB="true".'
|
||||
'Can not enable dynamic_eplb when DYNAMIC_EPLB is not set to "true" or "1".'
|
||||
):
|
||||
EPLBParamUtils.check_dynamic_eplb(True)
|
||||
|
||||
@@ -144,7 +144,7 @@ class TestEPLBParamUtils:
|
||||
with pytest.raises(
|
||||
ValueError,
|
||||
match=
|
||||
'Can not enable dynamic_eplb when not export DYNAMIC_EPLB="true".'
|
||||
'Can not enable dynamic_eplb when DYNAMIC_EPLB is not set to "true" or "1".'
|
||||
):
|
||||
EPLBParamUtils.check_dynamic_eplb(True)
|
||||
|
||||
@@ -152,6 +152,12 @@ class TestEPLBParamUtils:
|
||||
monkeypatch.setenv("DYNAMIC_EPLB", "true")
|
||||
EPLBParamUtils.check_dynamic_eplb(True)
|
||||
|
||||
monkeypatch.setenv("DYNAMIC_EPLB", "True")
|
||||
EPLBParamUtils.check_dynamic_eplb(True)
|
||||
|
||||
monkeypatch.setenv("DYNAMIC_EPLB", "1")
|
||||
EPLBParamUtils.check_dynamic_eplb(True)
|
||||
|
||||
def test_check_expert_map_path_none(self):
|
||||
EPLBParamUtils.check_expert_map_path(None)
|
||||
|
||||
|
||||
@@ -172,7 +172,10 @@ env_variables: Dict[str, Callable[[], Any]] = {
|
||||
lambda: int(os.getenv("VLLM_ASCEND_ENABLE_NZ", 1)),
|
||||
# Decide whether we should enable CP parallelism.
|
||||
"VLLM_ASCEND_ENABLE_CONTEXT_PARALLEL":
|
||||
lambda: bool(int(os.getenv("VLLM_ASCEND_ENABLE_CONTEXT_PARALLEL", '0')))
|
||||
lambda: bool(int(os.getenv("VLLM_ASCEND_ENABLE_CONTEXT_PARALLEL", '0'))),
|
||||
# Whether to anbale dynamic EPLB
|
||||
"DYNAMIC_EPLB":
|
||||
lambda: os.getenv("DYNAMIC_EPLB", "false").lower(),
|
||||
}
|
||||
|
||||
# end-env-vars-definition
|
||||
|
||||
@@ -22,6 +22,8 @@ import sys
|
||||
import torch
|
||||
from vllm.logger import logger
|
||||
|
||||
import vllm_ascend.envs as envs_ascend
|
||||
|
||||
|
||||
def determine_default_expert_map(global_expert_num, world_size, rank_id,
|
||||
global_redundant_expert_num):
|
||||
@@ -140,9 +142,10 @@ class EPLBParamUtils:
|
||||
return
|
||||
if not isinstance(dynamic_eplb, bool):
|
||||
raise TypeError("The dynamic_eplb is not bool.")
|
||||
if dynamic_eplb and os.getenv("DYNAMIC_EPLB", "false") != "true":
|
||||
|
||||
if dynamic_eplb and envs_ascend.DYNAMIC_EPLB not in ("true", "1"):
|
||||
raise ValueError(
|
||||
'Can not enable dynamic_eplb when not export DYNAMIC_EPLB="true".'
|
||||
'Can not enable dynamic_eplb when DYNAMIC_EPLB is not set to "true" or "1".'
|
||||
)
|
||||
|
||||
@staticmethod
|
||||
|
||||
@@ -16,11 +16,12 @@
|
||||
|
||||
import os
|
||||
|
||||
import vllm_ascend.envs as envs_ascend
|
||||
import vllm_ascend.patch.platform.patch_config # noqa
|
||||
import vllm_ascend.patch.platform.patch_distributed # noqa
|
||||
import vllm_ascend.patch.platform.patch_mamba_config # noqa
|
||||
import vllm_ascend.patch.platform.patch_sched_yield # noqa
|
||||
|
||||
if os.getenv("DYNAMIC_EPLB", "false") == "true" or os.getenv(
|
||||
if envs_ascend.DYNAMIC_EPLB not in ("true", "1") or os.getenv(
|
||||
"EXPERT_MAP_RECORD", "false") == "true":
|
||||
import vllm_ascend.patch.platform.patch_multiproc_executor # noqa
|
||||
|
||||
Reference in New Issue
Block a user