[Bugfix] Safeguard against missing backend in AttentionBackendEnum (vllm-project#28846)

jesse996 · gemini-code-assist[bot] · web-flow · commit 285eaa42857b · 2025-11-18T10:53:44.000Z
Signed-off-by: jesse &lt;szxfml@gmail.com&gt;
Signed-off-by: Song Zhixin &lt;szxfml@gmail.com&gt;
Co-authored-by: gemini-code-assist[bot] &lt;176961590+gemini-code-assist[bot]@users.noreply.github.com&gt;
diff --git a/vllm/attention/layer.py b/vllm/attention/layer.py
@@ -310,7 +310,8 @@ def __init__(
             kv_sharing_target_layer_name,
             **extra_impl_args,
         )
-        self.backend = AttentionBackendEnum[self.attn_backend.get_name()]
+        backend_name = self.attn_backend.get_name()
+        self.backend = AttentionBackendEnum.__members__.get(backend_name)
         self.dtype = dtype
 
         # For cuda-alike (CUDA and ROCM) and cpu platforms, we control how

Original file line number	Diff line number	Diff line change
`@@ -310,7 +310,8 @@ def __init__(`
`310`	`310`	`kv_sharing_target_layer_name,`
`311`	`311`	`**extra_impl_args,`
`312`	`312`	`)`
`313`		`- self.backend = AttentionBackendEnum[self.attn_backend.get_name()]`
	`313`	`+ backend_name = self.attn_backend.get_name()`
	`314`	`+ self.backend = AttentionBackendEnum.__members__.get(backend_name)`
`314`	`315`	`self.dtype = dtype`
`315`	`316`
`316`	`317`	`# For cuda-alike (CUDA and ROCM) and cpu platforms, we control how`