Account for SlidingWindowSpec in TPU worker

[vanbasten23]

Account for SlidingWindowSpec in TPU worker. Otherwise the TPU CI fails with an error

INFO 04-04 01:27:29 [loader.py:447] Loading weights took 2.43 seconds
--
  | ERROR 04-04 01:27:29 [core.py:390] EngineCore hit an exception: Traceback (most recent call last):
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/v1/engine/core.py", line 378, in run_engine_core
  | ERROR 04-04 01:27:29 [core.py:390]     engine_core = EngineCoreProc(*args, **kwargs)
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/v1/engine/core.py", line 319, in __init__
  | ERROR 04-04 01:27:29 [core.py:390]     super().__init__(vllm_config, executor_class, log_stats)
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/v1/engine/core.py", line 71, in __init__
  | ERROR 04-04 01:27:29 [core.py:390]     self._initialize_kv_caches(vllm_config)
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/v1/engine/core.py", line 132, in _initialize_kv_caches
  | ERROR 04-04 01:27:29 [core.py:390]     available_gpu_memory = self.model_executor.determine_available_memory()
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/v1/executor/abstract.py", line 66, in determine_available_memory
  | ERROR 04-04 01:27:29 [core.py:390]     output = self.collective_rpc("determine_available_memory")
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/executor/uniproc_executor.py", line 56, in collective_rpc
  | ERROR 04-04 01:27:29 [core.py:390]     answer = run_method(self.driver_worker, method, args, kwargs)
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/utils.py", line 2347, in run_method
  | ERROR 04-04 01:27:29 [core.py:390]     return func(*args, **kwargs)
  | ERROR 04-04 01:27:29 [core.py:390]   File "/workspace/vllm/vllm/v1/worker/tpu_worker.py", line 150, in determine_available_memory
  | ERROR 04-04 01:27:29 [core.py:390]     raise NotImplementedError
  | ERROR 04-04 01:27:29 [core.py:390] NotImplementedError
  | ERROR 04-04 01:27:29 [core.py:390]
  | CRITICAL 04-04 01:27:29 [core_client.py:361] Got fatal signal from worker processes, shutting down. See stack trace above for root cause

The not implemented case is SlidingWindowSpec.

Test plans:

pytest -v -s tests/entrypoints/llm/test_accuracy.py::test_lm_eval_accuracy_v1_engine
VLLM_USE_V1=1 python examples/offline_inference/tpu.py  # change the model to google/gemma-3-1b-it

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[robertgshaw2-redhat]

Thanks!