[Bugfix][HunyuanImage3.0] Fix KV reuse compatibility in SP scenarios (#3546)

Signed-off-by: dengyunyang <584797741@qq.com>

vllm_omni/diffusion/hooks/sequence_parallel.py

@@ -316,6 +316,13 @@ class SequenceParallelSplitHook(ModelHook):
             logger.warning_once(f"Expected tensor with {sp_input.expected_dims} dims, got {x.dim()}. Skipping split.")
             return x
 
+        split_dim = (
+            sp_input.split_dim if isinstance(sp_input, (SequenceParallelInput, SequenceParallelPartialInput)) else None
+        )
+        if split_dim is not None and x.size(split_dim) == 0:
+            logger.warning_once("Skip sharding for zero-sized tensors.")
+            return x
+
         def _raise_strict_divisibility_error(*, dim: int, seq_len: int, sp_size: int) -> None:
             # Keep message actionable: strict mode must be evenly shardable at the split hook level.
             msg = (