[bugfix] Fix SP + PP without specifying compile size

[angelayi]

Purpose

Fixes test failure in https://buildkite.com/vllm/ci/builds/35052/steps/canvas?sid=0199e88b-b61d-448b-a9cd-b96b4c779a60#0199e88b-b74c-49ca-8496-a34e8750a30a/217-11082

Test Plan

    def test_pp(self):
        config = CompilationConfig(
            mode=CompilationMode.VLLM_COMPILE,
            custom_ops=["+rms_norm"],
            pass_config=PassConfig(
                enable_async_tp=False,
                enable_sequence_parallelism=True,
                enable_noop=True,
            ),
            use_inductor_graph_partition=True,
        )

        llm = LLM(
            model="meta-llama/Llama-3.2-1B-Instruct",
            gpu_memory_utilization=0.6,
            max_model_len=2048, 
            max_num_seqs=8,
            compilation_config=config,
            tensor_parallel_size=2,  
            pipeline_parallel_size=2,  
            distributed_executor_backend="mp",
        )

        # Simple generation test
        prompts = ["Hello, my name is"]
        outputs = llm.generate(prompts, SamplingParams(temperature=0, max_tokens=32))

        # Print the outputs
        print("-" * 60)
        for output in outputs:
            prompt = output.prompt
            generated_text = output.outputs[0].text
            print(f"Prompt: {prompt!r}")
            print(f"Output: {generated_text!r}")
        print("-" * 60)

cc @ProExpertProg @cascade812