[ROCM] fix native attention function call (vllm-project#13650)

gongdao123 · shreyankg · commit 1e43d2e68c4c · 2025-05-02T23:47:45.000+05:30
diff --git a/vllm/attention/backends/rocm_flash_attn.py b/vllm/attention/backends/rocm_flash_attn.py
@@ -717,7 +717,6 @@ def forward(
                         self.num_heads,
                         self.head_size,
                         self.scale,
-                        causal_mask,
                         attn_masks,
                     )
                 else:
