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

[Bugfix] Properly set distributed_executor_backend in ParallelConfig (#4816)

Browse Source
This commit is contained in:
zifeitong
2024-05-15 07:22:09 -07:00
committed by GitHub
parent e9cdd2b1e2
commit a5675d348b
2 changed files with 8 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
vllm/engine/arg_utils.py
View File

@@ -548,14 +548,18 @@ class EngineArgs:
model_config.get_sliding_window(),
self.enable_prefix_caching)
parallel_config = ParallelConfig(
self.pipeline_parallel_size, self.tensor_parallel_size,
self.worker_use_ray, self.max_parallel_loading_workers,
self.pipeline_parallel_size,
self.tensor_parallel_size,
self.worker_use_ray,
self.max_parallel_loading_workers,
self.disable_custom_all_reduce,
TokenizerPoolConfig.create_config(
self.tokenizer_pool_size,
self.tokenizer_pool_type,
self.tokenizer_pool_extra_config,
), self.ray_workers_use_nsight)
),
self.ray_workers_use_nsight,
distributed_executor_backend=self.distributed_executor_backend)
speculative_config = SpeculativeConfig.maybe_create_spec_config(
target_model_config=model_config,