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, 18:30 UTC
[V1][Bugfix] Fix oracle for device checking (#15104)
Signed-off-by: Roger Wang <ywang@roblox.com>
Roger Wang committed
Mar 19, 2025 at 03:35 UTC
dafb4e504ad1b83320bc2800675b52be0324ba12
1 file changed
+7
vllm/engine/arg_utils.py
+7
@@ -1578,6 +1578,13 @@ class EngineArgs:
1578
_raise_or_fallback(feature_name=name, recommend_to_remove=True)
1579
return False
1580
1581
+ # No support for device type other than CUDA, AMD (experiemntal) or
1582
+ # TPU (experimental) so far.
1583
+ if not (current_platform.is_cuda_alike() or current_platform.is_tpu()):
1584
+ _raise_or_fallback(
1585
+ feature_name=f"device type={current_platform.device_type}",
1586
+ recommend_to_remove=False)
1587
+ return False
1588
#############################################################
1589
# Experimental Features - allow users to opt in.
1590