mirror of
https://github.com/wassname/vllm.git
synced 2026-06-27 18:27:02 +08:00
Add warning
This commit is contained in:
@@ -19,6 +19,9 @@ class CacheEngine:
|
||||
num_cpu_blocks: int,
|
||||
dtype: torch.dtype,
|
||||
) -> None:
|
||||
if head_size % 16 != 0:
|
||||
raise ValueError(f'head_size ({head_size}) must be a multiple of 16.')
|
||||
|
||||
self.worker_id = worker_id
|
||||
self.gpu_id = gpu_id
|
||||
self.num_layers = num_layers
|
||||
|
||||
Reference in New Issue
Block a user