diff --git a/tests/models/multimodal/conftest.py b/tests/models/multimodal/conftest.py
index 4243298cdc896..31d99218c8276 100644
--- a/tests/models/multimodal/conftest.py
+++ b/tests/models/multimodal/conftest.py
@@ -19,7 +19,7 @@ def pytest_collection_modifyitems(config, items):
         return
 
     # Disable Flash/MemEfficient SDP on ROCm to avoid HF Transformers
-    # accuracy issues
+    # accuracy issues: https://github.com/vllm-project/vllm/issues/30167
     # TODO: Remove once ROCm SDP accuracy issues are resolved on HuggingFace
     torch.backends.cuda.enable_flash_sdp(False)
     torch.backends.cuda.enable_mem_efficient_sdp(False)
diff --git a/vllm/attention/layers/mm_encoder_attention.py b/vllm/attention/layers/mm_encoder_attention.py
index 25f54cc867b5a..1c1623b13f55a 100644
--- a/vllm/attention/layers/mm_encoder_attention.py
+++ b/vllm/attention/layers/mm_encoder_attention.py
@@ -136,7 +136,7 @@ class MMEncoderAttention(CustomOp):
             cu_seqlens=cu_seqlens,
         )
         if is_reshaped:
-            output = output.view(bsz, q_len, -1)
+            output = output.reshape(bsz, q_len, -1)
         return output
 
     def _forward_fa(
@@ -174,7 +174,7 @@ class MMEncoderAttention(CustomOp):
             fa_version=self._fa_version,
         )
         if is_reshaped:
-            output = output.view(bsz, q_len, -1)
+            output = output.reshape(bsz, q_len, -1)
         return output
 
     def forward_native(