[Fix] Restore MammothModa2 support

Signed-off-by: hsliuustc0106 <hsliuustc0106@users.noreply.github.com>

vllm_omni/worker/gpu_generation_model_runner.py

@@ -792,9 +792,9 @@ class GPUGenerationModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin
                 input_ids = self.input_ids.gpu[:num_tokens_padded]
                 inputs_embeds = None
 
-            # Some generation-stage models require model-specific runtime
-            # information (such as image size and conditioning embeddings)
-            # even during the dummy profiling run that vLLM uses to
+            # Some generation-stage models (e.g. MammothModa2DiTPipeline) require
+            # model-specific runtime information (such as image size and conditioning
+            # embeddings) even during the dummy profiling run that vLLM uses to
             # estimate KV-cache capacity.  get_dummy_runtime_additional_information
             # provides placeholder values of the correct shape so that the profiling
             # run does not raise an error due to missing inputs.