prefill only attention

vllm-project · Oct 7, 2024 · 0333601 · 0333601
1 parent 748fdee
commit 0333601
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/vllm/attention/layer.py b/vllm/attention/layer.py
@@ -81,7 +81,7 @@ def __init__(
             dtype = torch.get_default_dtype()
             self.attn_backend = get_attn_backend(
                 num_heads, head_size, num_kv_heads, sliding_window, dtype,
-                kv_cache_dtype, block_size, blocksparse_params is not None)
+                kv_cache_dtype, block_size, blocksparse_params is not None)()
         else:
             self.attn_backend = attn_backend