A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 13:30 UTC
[Misc] Clean up redundant code in uniproc_executor.py (#16762)
Signed-off-by: Lifu Huang <lifu.hlf@gmail.com>
Lifu Huang committed
Apr 24, 2025 at 22:49 UTC
5aa6efb9a5929d9b4b6b2c231b81f63587e83c77
1 file changed
+2
-2
vllm/executor/uniproc_executor.py
+2
-2
@@ -34,13 +34,13 @@ class UniProcExecutor(ExecutorBase):
34
if len(device_info) > 1:
35
local_rank = int(device_info[1])
36
rank = 0
37
+ is_driver_worker = True
38
kwargs = dict(
39
vllm_config=self.vllm_config,
40
local_rank=local_rank,
41
rank=rank,
42
distributed_init_method=distributed_init_method,
42
- is_driver_worker=(not self.parallel_config)
43
- or (rank % self.parallel_config.tensor_parallel_size == 0),
43
+ is_driver_worker=is_driver_worker,
44
)
45
self.collective_rpc("init_worker", args=([kwargs], ))
46
self.collective_rpc("init_device")