diff --git a/vllm/compilation/decorators.py b/vllm/compilation/decorators.py
index 0474db0820c73..5f1b268a1d6fe 100644
--- a/vllm/compilation/decorators.py
+++ b/vllm/compilation/decorators.py
@@ -170,7 +170,7 @@ def _support_torch_compile(
         # e.g. TPU has the compilation logic in model runner, so we don't
         # need to compile the model inside.
         if self.do_not_compile or torch.compiler.is_compiling():
-            logger.info("SKIPPING COMPILATION")
+            # logger.info("SKIPPING COMPILATION")
             return self.forward(*args, **kwargs)
 
         # the first compilation needs to have dynamic shapes marked
diff --git a/vllm/v1/attention/backends/flash_attn.py b/vllm/v1/attention/backends/flash_attn.py
index 88be86749d8d3..788aa0c537af5 100755
--- a/vllm/v1/attention/backends/flash_attn.py
+++ b/vllm/v1/attention/backends/flash_attn.py
@@ -337,14 +337,6 @@ class FlashAttentionMetadataBuilder(
         # populated on first build() call.
         self.aot_sliding_window: Optional[tuple[int, int]] = None
 
-    def build(
-        self, common_prefix_len: int,
-        common_attn_metadata: CommonAttentionMetadata
-    ) -> FlashAttentionMetadata:
-        num_reqs = common_attn_metadata.num_reqs
-        num_actual_tokens = common_attn_metadata.num_actual_tokens
-        max_query_len = common_attn_metadata.max_query_len
-
     def build_slice(
         self,
         req_slice: slice,
diff --git a/vllm/v1/attention/backends/mla/common.py b/vllm/v1/attention/backends/mla/common.py
index fbd95423b1c3b..fa9e94ad189ac 100644
--- a/vllm/v1/attention/backends/mla/common.py
+++ b/vllm/v1/attention/backends/mla/common.py
@@ -610,7 +610,7 @@ class MLACommonMetadataBuilder(AttentionMetadataBuilder[M]):
         num_actual_tokens = common_attn_metadata.num_actual_tokens
         max_query_len = common_attn_metadata.max_query_len
 
-        assert self._num_decodes + self._num_prefills == num_reqs
+        # assert self._num_decodes + self._num_prefills == num_reqs
         return self.build_slice(
             req_slice=slice(0, num_reqs),
             token_slice=slice(0, num_actual_tokens),