From 9854f2022515e6ef191b672f2b2470b4f01d82ba Mon Sep 17 00:00:00 2001 From: drbh Date: Wed, 12 Jun 2024 17:12:18 +0000 Subject: [PATCH] fix: set sharded true if WORLD_SIZE is set --- server/text_generation_server/cli.py | 3 +++ 1 file changed, 3 insertions(+) diff --git a/server/text_generation_server/cli.py b/server/text_generation_server/cli.py index 430323bc..1f5f3232 100644 --- a/server/text_generation_server/cli.py +++ b/server/text_generation_server/cli.py @@ -44,6 +44,9 @@ def serve( otlp_endpoint: Optional[str] = None, max_input_tokens: Optional[int] = None, ): + # derive sharded from environment variables if not provided + sharded = sharded or os.getenv("WORLD_SIZE", None) is not None + if sharded: assert ( os.getenv("RANK", None) is not None