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, 09:45 UTC
[Misc][MacOS] fix bfloat16 error (#18249)
Signed-off-by: reidliu41 <reid201711@gmail.com> Co-authored-by: reidliu41 <reid201711@gmail.com>
Reid committed
May 16, 2025 at 23:05 UTC
d3d91b6f714184ef387701bdc8cfd5990d9e78bb
1 file changed
+1
-1
vllm/platforms/cpu.py
+1
-1
@@ -36,7 +36,7 @@ class CpuPlatform(Platform):
36
# instead of checking the OS. For instance M2 shall supports bf16
37
# already. But we need to modify `cpu_extension.cmake` to activate
38
# the feature in the build.
39
- return [torch.bfloat16, torch.float32]
39
+ return [torch.float16, torch.float32]
40
# x86/aarch64 CPU has supported both bf16 and fp16 natively.
41
return [torch.bfloat16, torch.float16, torch.float32]
42