remove torch sdpa as input

Fridah-nv · Fridah-nv · commit 985fb449c51c · 2025-09-26T21:49:20.000Z
Signed-off-by: Frida Hou &lt;201670829+Fridah-nv@users.noreply.github.com&gt;
diff --git a/tensorrt_llm/_torch/auto_deploy/transform/library/attention.py b/tensorrt_llm/_torch/auto_deploy/transform/library/attention.py
@@ -529,7 +529,6 @@ def _apply(
 
         # List of SDPA operations to look for
         sdpa_ops = {
-            torch.ops.auto_deploy.torch_attention_sdpa,
             torch.ops.auto_deploy.torch_attention_grouped_sdpa,
         }
 
diff --git a/tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.py b/tests/unittest/_torch/auto_deploy/unit/singlegpu/transformations/library/test_attention_matcher.py
@@ -1157,7 +1157,6 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 @pytest.mark.parametrize(
     "model_config",
     [
-        {"type": "standard", "use_grouped_sdpa": False, "name": "SDPA"},
         {"type": "standard", "use_grouped_sdpa": True, "name": "GroupedSDPA"},
         {"type": "already_bsnd", "name": "DirectBSND"},
     ],

Original file line number	Diff line number	Diff line change
`@@ -529,7 +529,6 @@ def _apply(`
`529`	`529`
`530`	`530`	`# List of SDPA operations to look for`
`531`	`531`	`sdpa_ops = {`
`532`		`- torch.ops.auto_deploy.torch_attention_sdpa,`
`533`	`532`	`torch.ops.auto_deploy.torch_attention_grouped_sdpa,`
`534`	`533`	`}`
`535`	`534`
Original file line number	Diff line number	Diff line change
`@@ -1157,7 +1157,6 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:`
`1157`	`1157`	`@pytest.mark.parametrize(`
`1158`	`1158`	`"model_config",`
`1159`	`1159`	`[`
`1160`		`- {"type": "standard", "use_grouped_sdpa": False, "name": "SDPA"},`
`1161`	`1160`	`{"type": "standard", "use_grouped_sdpa": True, "name": "GroupedSDPA"},`
`1162`	`1161`	`{"type": "already_bsnd", "name": "DirectBSND"},`
`1163`	`1162`	`],`