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 17, 03:00 UTC
[Core] fix custom allreduce default value (#4040)
youkaichao committed
Apr 12, 2024 at 16:40 UTC
fbb9d9eef48a29e0ea821bbf399e4bf9a08d6ac1
1 file changed
+1
-1
vllm/entrypoints/llm.py
+1
-1
@@ -86,7 +86,7 @@ class LLM:
86
swap_space: int = 4,
87
enforce_eager: bool = False,
88
max_context_len_to_capture: int = 8192,
89
- disable_custom_all_reduce: bool = True,
89
+ disable_custom_all_reduce: bool = False,
90
**kwargs,
91
) -> None:
92
if "disable_log_stats" not in kwargs: