[XPU][CI] enhance xpu test support

tests/conftest.py

@@ -759,7 +759,8 @@ class VllmRunner:
     - `trust_remote_code`: Set to `True` instead of `False` for convenience.
     - `seed`: Set to `0` instead of `None` for test reproducibility.
     - `max_model_len`: Set to `1024` instead of `None` to reduce memory usage.
-    - `block_size`: Set to `16` instead of `None` to reduce memory usage.
+    - `block_size`: To reduce memory usage, set default to `64` if on XPU devices,
+      otherwise default to `16`.
     - `enable_chunked_prefill`: Set to `False` instead of `None` for
       test reproducibility.
     - `enforce_eager`: Set to `False` to test CUDA graph.
@@ -777,7 +778,7 @@ def __init__(
         dtype: str = "auto",
         disable_log_stats: bool = True,
         tensor_parallel_size: int = 1,
-        block_size: int = 16,
+        block_size: int = 16 if not hasattr(torch, 'xpu') else 64,
         enable_chunked_prefill: Optional[bool] = False,
         swap_space: int = 4,
         enforce_eager: Optional[bool] = False,

tests/utils.py

@@ -730,6 +730,11 @@ def fork_new_process_for_each_test(
 
     @functools.wraps(f)
     def wrapper(*args: _P.args, **kwargs: _P.kwargs) -> None:
+        # To use XPU with multiprocessing, must use the 'spawn' start method via 'VLLM_WORKER_MULTIPROC_METHOD=spawn'
+        if current_platform.is_xpu():
+            f(*args, **kwargs)
+            return
+
         # Make the process the leader of its own process group
         # to avoid sending SIGTERM to the parent process
         os.setpgrp()
@@ -817,7 +822,7 @@ def create_new_process_for_each_test(
     """Creates a decorator that runs each test function in a new process.
 
     Args:
-        method: The process creation method. Can be either "spawn" or "fork". 
+        method: The process creation method. Can be either "spawn" or "fork".
                If not specified,
                it defaults to "spawn" on ROCm platforms and "fork" otherwise.
 

vllm/distributed/device_communicators/xpu_communicator.py

@@ -53,3 +53,8 @@ def gather(self,
         else:
             output_tensor = None
         return output_tensor
+
+    def broadcast(self,
+                  input_: torch.Tensor,
+                  src: int = 0) -> None:
+        dist.broadcast(input_, src=src, group=self.device_group)

vllm/distributed/parallel_state.py

@@ -240,6 +240,8 @@ def __init__(
 
         if current_platform.is_cuda_alike():
             self.device = torch.device(f"cuda:{local_rank}")
+        elif current_platform.is_xpu():
+            self.device = torch.device(f"xpu:{local_rank}")
         elif current_platform.is_out_of_tree():
             self.device = torch.device(
                 f"{current_platform.device_name}:{local_rank}")
@@ -1317,13 +1319,13 @@ def in_the_same_node_as(pg: Union[ProcessGroup, StatelessProcessGroup],
 
 def is_global_first_rank() -> bool:
     """
-    Check if the current process is the first rank globally across all 
+    Check if the current process is the first rank globally across all
     parallelism strategies (PP, TP, DP, EP, etc.).
-    
+
     Unlike group-specific checks like `get_tensor_model_parallel_rank() == 0`
     or `get_pp_group().is_first_rank`, this function checks the global rank
     across all parallelism dimensions.
-    
+
     Returns:
         bool: True if this is the global first rank (rank 0), False otherwise.
               Returns True if distributed is not initialized (single process).
@@ -1352,7 +1354,7 @@ def _node_count(pg: Union[ProcessGroup, StatelessProcessGroup]) -> int:
 
     Args:
         pg: The process group to analyze
-        
+
     Returns:
         int: The total number of nodes
     """

vllm/platforms/xpu.py

@@ -78,6 +78,14 @@ def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
         if cache_config and cache_config.block_size is None:
             cache_config.block_size = 64
 
+        # FIXME: Temporarily forcing eager mode
+        # remove after t.compile support stabilizes.
+        if envs.VLLM_USE_V1 and vllm_config.model_config is not None and \
+            not vllm_config.model_config.enforce_eager:
+            from vllm.config import CompilationLevel
+            vllm_config.compilation_config.level = \
+                CompilationLevel.NO_COMPILATION
+
         # Instances created using VllmConfig() typically have model_config as
         # None by default. The modification involves adding a check to prevent
         # potential null exceptions check and update model config.
@@ -93,9 +101,6 @@ def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
                     "mode.")
                 model_config.enforce_eager = True
 
-        if vllm_config.device_config is not None:
-            assert vllm_config.device_config.device_type == "xpu"
-
         # check and update parallel config
         parallel_config = vllm_config.parallel_config
         parallel_config.worker_cls = "vllm.v1.worker.xpu_worker.XPUWorker"
@@ -114,7 +119,8 @@ def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
                 logger.warning(
                     "Please use spawn as start method if you want to use mp.")
         elif parallel_config.distributed_executor_backend != "ray" and \
-                parallel_config.distributed_executor_backend != "uni":
+                parallel_config.distributed_executor_backend != "uni" and \
+                parallel_config.distributed_executor_backend != "external_launcher":
             logger.warning(
                 "%s is not supported on XPU, fallback to ray distributed"
                 " executor backend.",

vllm/v1/worker/xpu_model_runner.py

@@ -27,7 +27,7 @@ def __init__(
         self.cascade_attn_enabled = False
 
     def _init_device_properties(self) -> None:
-        pass
+        self.num_sms = None
 
     def _sync_device(self) -> None:
         torch.xpu.synchronize()