xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-16 14:27:19 +08:00
Code Issues Packages Projects Releases Wiki Activity

updated

Browse Source 
Signed-off-by: Robert Shaw <robshaw@redhat.com>
This commit is contained in:
Robert Shaw 2025-07-15 02:09:43 +00:00
parent e830434fe2
commit 9d762c3aa5
2 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/layers/fused_moe/pplx_prepare_finalize.py
View File

@ -2,6 +2,7 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from typing import Optional from typing import Optional
import os
import time import time
import pplx_kernels as pplx import pplx_kernels as pplx
import torch import torch
@ -210,6 +211,7 @@ class PplxPrepareAndFinalize(mk.FusedMoEPrepareAndFinalize):
bound_m=bound_m, bound_m=bound_m,
) )
end = time.perf_counter() end = time.perf_counter()
if os.getenv("LOG_TIME") == "1":
logger.info("dispatch took %.3f ms", (end - start) * 1000) logger.info("dispatch took %.3f ms", (end - start) * 1000)
if expert_x_scale is not None: if expert_x_scale is not None:
@ -259,4 +261,5 @@ class PplxPrepareAndFinalize(mk.FusedMoEPrepareAndFinalize):
expert_y=fused_expert_output, expert_y=fused_expert_output,
bound_m=bound_m) bound_m=bound_m)
end = time.perf_counter() end = time.perf_counter()
if os.getenv("LOG_TIME") == "1":
logger.info("combine took %.3f ms", (end - start) * 1000) logger.info("combine took %.3f ms", (end - start) * 1000)

1
vllm/v1/engine/core.py
View File

@ -946,7 +946,6 @@ class DPEngineCoreProc(EngineCoreProc):
# We are in a running state and so must execute a dummy pass # We are in a running state and so must execute a dummy pass
# if the model didn't execute any ready requests. # if the model didn't execute any ready requests.
logger.info("Executing dummy batch for wave %d.", self.current_wave)
self.execute_dummy_batch() self.execute_dummy_batch()
# 3) All-reduce operation to determine global unfinished reqs. # 3) All-reduce operation to determine global unfinished reqs.