[TPU] add tpu_inference (vllm-project#27277)

Signed-off-by: Johnny Yang <johnnyyang@google.com>

Author: jcyang43

requirements/tpu.txt

@@ -12,6 +12,4 @@ ray[data]
 setuptools==78.1.0
 nixl==0.3.0
 tpu_info==0.4.0
-
-# Install torch_xla
-torch_xla[tpu, pallas]==2.8.0
+tpu-inference==0.11.1

vllm/distributed/device_communicators/tpu_communicator.py

@@ -97,11 +97,3 @@ def all_reduce(self, input_: torch.Tensor) -> torch.Tensor:
     def all_gather(self, input_: torch.Tensor, dim: int = -1) -> torch.Tensor:
         assert dim == -1, "TPUs only support dim=-1 for all-gather."
         return xm.all_gather(input_, dim=dim)
-
-
-if USE_TPU_INFERENCE:
-    from tpu_inference.distributed.device_communicators import (
-        TpuCommunicator as TpuInferenceCommunicator,
-    )
-
-    TpuCommunicator = TpuInferenceCommunicator  # type: ignore

vllm/platforms/tpu.py

@@ -267,7 +267,9 @@ def check_max_model_len(cls, max_model_len: int) -> int:
 
 
 try:
-    from tpu_inference.platforms import TpuPlatform as TpuInferencePlatform
+    from tpu_inference.platforms.tpu_platforms import (
+        TpuPlatform as TpuInferencePlatform,
+    )
 
     TpuPlatform = TpuInferencePlatform  # type: ignore
     USE_TPU_INFERENCE = True

vllm/v1/worker/tpu_worker.py

@@ -346,6 +346,6 @@ def apply_model(self, fn: Callable[[nn.Module], _R]) -> _R:
 
 
 if USE_TPU_INFERENCE:
-    from tpu_inference.worker import TPUWorker as TpuInferenceWorker
+    from tpu_inference.worker.tpu_worker import TPUWorker as TpuInferenceWorker
 
     TPUWorker = TpuInferenceWorker  # type: ignore