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, 04:45 UTC
[Kernel] mark TorchSDPABackend swap_blocks NotImplementedError (#19749)
Ning Xie committed
Jun 21, 2025 at 02:18 UTC
71baf85ae11be24d4ea32d30cb5b8dfb0912a6cc
1 file changed
+1
-1
vllm/attention/backends/torch_sdpa.py
+1
-1
@@ -65,7 +65,7 @@ class TorchSDPABackend(AttentionBackend):
65
dst_kv_cache: torch.Tensor,
66
src_to_dst: torch.Tensor,
67
) -> None:
68
- PagedAttention.swap_blocks(src_kv_cache, dst_kv_cache, src_to_dst)
68
+ raise NotImplementedError("Swap is not supported in TorchSDPABackend.")
69
70
@staticmethod
71
def copy_blocks(