xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-20 18:27:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

Correct eplb state logs

Browse Source 
Signed-off-by: ilmarkov <markovilya197@gmail.com>
This commit is contained in:
ilmarkov 2025-12-11 09:20:33 +00:00
parent 6b2a1de500
commit fc54d760a6
1 changed files with 14 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
vllm/distributed/eplb/eplb_state.py
View File

@ -27,7 +27,6 @@ physical experts.
""" """
import threading import threading
import time
from collections.abc import Sequence from collections.abc import Sequence
from dataclasses import dataclass from dataclasses import dataclass
@ -699,11 +698,14 @@ class EplbState:
ep_group = get_ep_group().device_group ep_group = get_ep_group().device_group
ep_rank = ep_group.rank() ep_rank = ep_group.rank()
time_start = None start_event = None
end_event = None
is_main_rank = ep_rank == 0 is_main_rank = ep_rank == 0
if is_main_rank: if is_main_rank:
torch.cuda.synchronize() if not self.is_async or is_profile:
time_start = time.perf_counter() start_event = torch.cuda.Event(enable_timing=True)
end_event = torch.cuda.Event(enable_timing=True)
start_event.record()
logger.info( logger.info(
"Rearranging experts %s %s...", "Rearranging experts %s %s...",
"(async mode)" if self.is_async else "sync mode", "(async mode)" if self.is_async else "sync mode",
@ -864,13 +866,15 @@ class EplbState:
new_logical_replica_count new_logical_replica_count
) )
if is_main_rank: if is_main_rank:
assert time_start is not None assert start_event is not None
torch.cuda.synchronize() assert end_event is not None
time_end = time.perf_counter() end_event.record()
end_event.synchronize()
gpu_elapsed = start_event.elapsed_time(end_event) / 1000.0
logger.info( logger.info(
"Rearranged experts%sin %.2f seconds.", "Rearranged experts %s in %.2f s.",
" (profile) " if is_profile else " ", " (profile) " if is_profile else " ",
time_end - time_start, gpu_elapsed,
) )
else: else:
max_slots = eplb_model_state.logical_to_physical_map.shape[-1] max_slots = eplb_model_state.logical_to_physical_map.shape[-1]
@ -1010,7 +1014,7 @@ class EplbState:
# After the main thread consumes, advance layer_to_transfer # After the main thread consumes, advance layer_to_transfer
model_state.layer_to_transfer += 1 model_state.layer_to_transfer += 1
model_state.ep_buffer_ready = 0 model_state.ep_buffer_ready = 0
logger.info( logger.debug(
"model %s successfully move_to_workspace layer %d", "model %s successfully move_to_workspace layer %d",
model_state.model_name, model_state.model_name,
transferred_layer, transferred_layer,