@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / d53575a5f0
Fork 0

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 16, 03:15 UTC

[ci] fix gh200 tests (#11919)

Signed-off-by: youkaichao <youkaichao@gmail.com>

youkaichao committed Jan 10, 2025 at 16:25 UTC d53575a5f0e5c0f9003b4ec6e33c8bf761e93cef
1 file changed +3 -1
vllm/model_executor/model_loader/weight_utils.py
+3 -1
@@ -29,7 +29,9 @@ from vllm.utils import PlaceholderModule
29
30 try:
31 from runai_model_streamer import SafetensorsStreamer
32 -except ImportError:
32 +except (ImportError, OSError):
33 + # see https://github.com/run-ai/runai-model-streamer/issues/26
34 + # OSError will be raised on arm64 platform
35 runai_model_streamer = PlaceholderModule(
36 "runai_model_streamer") # type: ignore[assignment]
37 SafetensorsStreamer = runai_model_streamer.placeholder_attr(