Skip to content

Commit 9252ae5

Browse files
committed
fix tp
Signed-off-by: Lu Fang <fanglu@fb.com>
1 parent 713c933 commit 9252ae5

File tree

4 files changed

+141
-89
lines changed

4 files changed

+141
-89
lines changed

vllm/v1/engine/core.py

Lines changed: 1 addition & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -43,9 +43,6 @@
4343
from vllm.v1.serial_utils import MsgpackDecoder, MsgpackEncoder
4444
from vllm.v1.structured_output import StructuredOutputManager
4545
from vllm.version import __version__ as VLLM_VERSION
46-
from vllm.v1.worker.intermediates_logging import (
47-
intermediate_logging
48-
)
4946

5047
logger = init_logger(__name__)
5148

@@ -221,8 +218,7 @@ def abort_requests(self, request_ids: list[str]):
221218
def execute_model(self, scheduler_output: SchedulerOutput):
222219
try:
223220
# Execute the model
224-
with intermediate_logging(self.vllm_config.il_config):
225-
return self.model_executor.execute_model(scheduler_output)
221+
return self.model_executor.execute_model(scheduler_output)
226222
except Exception as err:
227223
# We do not want to catch BaseException here since we're only
228224
# interested in dumping info when the exception is due to an

vllm/v1/worker/gpu_worker.py

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -30,6 +30,7 @@
3030
from vllm.v1.utils import report_usage_stats
3131
from vllm.v1.worker.gpu_model_runner import GPUModelRunner
3232
from vllm.v1.worker.worker_base import WorkerBase
33+
from vllm.v1.worker.intermediates_logging import intermediate_logging
3334

3435
logger = init_logger(__name__)
3536

@@ -320,8 +321,9 @@ def execute_model(
320321
get_pp_group().recv_tensor_dict(
321322
all_gather_group=get_tp_group()))
322323

323-
output = self.model_runner.execute_model(scheduler_output,
324-
intermediate_tensors)
324+
with intermediate_logging(self.vllm_config.il_config):
325+
output = self.model_runner.execute_model(scheduler_output,
326+
intermediate_tensors)
325327

326328
parallel_config = self.vllm_config.parallel_config
327329
if parallel_config.distributed_executor_backend != "external_launcher" \

0 commit comments

Comments
 (0)