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

[Config] Clean up SchedulerConfig initialization (#28665)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-11-14 22:41:02 +08:00
committed by GitHub
parent 96b23b8e3b
commit 511a6b611d
9 changed files with 182 additions and 163 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/platforms/cpu.py
View File

@@ -15,7 +15,6 @@ import torch
from vllm import envs
from vllm.logger import init_logger
from vllm.utils import DEFAULT_MAX_NUM_BATCHED_TOKENS
from .interface import CpuArchEnum, Platform, PlatformEnum
@@ -339,10 +338,9 @@ class CpuPlatform(Platform):
"prefill and prefix caching to be disabled."
)
vllm_config.scheduler_config.enable_chunked_prefill = False
vllm_config.scheduler_config.chunked_prefill_enabled = False
vllm_config.scheduler_config.max_num_batched_tokens = max(
vllm_config.scheduler_config.max_model_len,
DEFAULT_MAX_NUM_BATCHED_TOKENS,
vllm_config.scheduler_config.DEFAULT_MAX_NUM_BATCHED_TOKENS,
)
@classmethod

4
vllm/platforms/tpu.py
View File

@@ -10,7 +10,6 @@ from tpu_info import device
from vllm.inputs import ProcessorInputs, PromptType
from vllm.logger import init_logger
from vllm.sampling_params import SamplingParams, SamplingType
from vllm.utils import DEFAULT_MAX_NUM_BATCHED_TOKENS
from .interface import Platform, PlatformEnum
@@ -186,10 +185,9 @@ class TpuPlatform(Platform):
"prefill and prefix caching to be disabled."
)
vllm_config.scheduler_config.enable_chunked_prefill = False
vllm_config.scheduler_config.chunked_prefill_enabled = False
vllm_config.scheduler_config.max_num_batched_tokens = max(
vllm_config.scheduler_config.max_model_len,
DEFAULT_MAX_NUM_BATCHED_TOKENS,
vllm_config.scheduler_config.DEFAULT_MAX_NUM_BATCHED_TOKENS,
)
@classmethod

4
vllm/platforms/xpu.py
View File

@@ -9,7 +9,6 @@ import torch
import vllm.envs as envs
from vllm.logger import init_logger
from vllm.utils import DEFAULT_MAX_NUM_BATCHED_TOKENS
from .interface import DeviceCapability, Platform, PlatformEnum
@@ -185,10 +184,9 @@ class XPUPlatform(Platform):
"prefill and prefix caching to be disabled."
)
vllm_config.scheduler_config.enable_chunked_prefill = False
vllm_config.scheduler_config.chunked_prefill_enabled = False
vllm_config.scheduler_config.max_num_batched_tokens = max(
vllm_config.scheduler_config.max_model_len,
DEFAULT_MAX_NUM_BATCHED_TOKENS,
vllm_config.scheduler_config.DEFAULT_MAX_NUM_BATCHED_TOKENS,
)
@classmethod