diff --git a/specforge/algorithms/common/dflash_family_model.py b/specforge/algorithms/common/dflash_family_model.py
index 1f3106517..2c0bdfe62 100644
--- a/specforge/algorithms/common/dflash_family_model.py
+++ b/specforge/algorithms/common/dflash_family_model.py
@@ -25,7 +25,6 @@
 if hasattr(torch, "npu") and torch.npu.is_available():
     FLEX_ATTENTION_AVAILABLE = False
 
-
 _VALID_LOSS_TYPES = {
     "dflash",
     "dpace",
@@ -363,9 +362,20 @@ def _forward_draft_blocks(
             # DFlash's dynamic short-query batches are training/prefill shaped,
             # not autoregressive decoding.  AUTO may route q_len < 128 to the
             # more restrictive flex-decoding kernel, whose config set can be
-            # empty for DFlash's sparse BlockMask.  Keep the general Triton
+            # empty for DFlash's sparse BlockMask.  Force the general Triton
             # Flex Attention kernel for every DFlash-family batch.
-            draft_kwargs["kernel_options"] = {"BACKEND": "TRITON"}
+            #
+            # The "BACKEND" kernel_option only exists on torch >= 2.11, where
+            # the inductor lowering sanitizes it out of the generated Triton
+            # constexprs.  On older builds (including current torch ROCm wheels)
+            # the string leaks into the kernel as a bare identifier and fails to
+            # compile (NameError: 'TRITON' is not defined), so we fall back to
+            # FORCE_USE_FLEX_ATTENTION, which selects the same kernel and has
+            # been supported since torch 2.5.
+            if torch.__version__ >= "2.11":
+                draft_kwargs["kernel_options"] = {"BACKEND": "TRITON"}
+            else:
+                draft_kwargs["kernel_options"] = {"FORCE_USE_FLEX_ATTENTION": True}
         output_hidden = self.draft_model(
             position_ids=full_position_ids,
             noise_embedding=noise_embedding,
diff --git a/tests/test_modeling/test_dflash_sliding.py b/tests/test_modeling/test_dflash_sliding.py
index 34cbc1487..b679ebf22 100644
--- a/tests/test_modeling/test_dflash_sliding.py
+++ b/tests/test_modeling/test_dflash_sliding.py
@@ -169,7 +169,12 @@ def test_online_wrapper_forces_standard_triton_flex_backend(self):
                 loss_mask=torch.ones(1, 4),
             )
 
-        self.assertEqual(layers[0].kernel_options, {"BACKEND": "TRITON"})
+        expected_kernel_options = (
+            {"BACKEND": "TRITON"}
+            if torch.__version__ >= "2.11"
+            else {"FORCE_USE_FLEX_ATTENTION": True}
+        )
+        self.assertEqual(layers[0].kernel_options, expected_kernel_options)
 
 
 class TestDFlashSlidingConfig(unittest.TestCase):