Convert formatting to use ruff instead of yapf + isort (#26247)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 07:06:22 -07:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115205 additions and 94107 deletions
+7 -12
View File
@@ -19,7 +19,6 @@ logger = init_logger(__name__)
class CPUModelRunner(GPUModelRunner):
def __init__(self, vllm_config: VllmConfig, device: torch.device):
with _torch_cuda_wrapper():
super().__init__(vllm_config, device)
@@ -35,14 +34,15 @@ class CPUModelRunner(GPUModelRunner):
# Note: Remove the override after new attention backend finished
def _may_reorder_batch(self, scheduler_output: "SchedulerOutput") -> None:
if len(self.kv_cache_config.kv_cache_groups) > 1:
raise ValueError("Multiple KVCacheGroups is not"
"currently supported with CPU model runner.")
raise ValueError(
"Multiple KVCacheGroups is not"
"currently supported with CPU model runner."
)
super()._may_reorder_batch(scheduler_output)
def _postprocess_tensors(self) -> None:
# Note: replace device tensors with cpu tensors
def replace_tensor(obj: Any, cpu_attr_name: str,
device_attr_name) -> None:
def replace_tensor(obj: Any, cpu_attr_name: str, device_attr_name) -> None:
cpu_tensor = getattr(obj, cpu_attr_name, None)
device_tensor = getattr(obj, device_attr_name, None)
if cpu_tensor is not None and device_tensor is not None:
@@ -68,8 +68,7 @@ class CPUModelRunner(GPUModelRunner):
self.model = get_model(vllm_config=self.vllm_config)
if self.lora_config:
self.model = self.load_lora_model(self.model, self.vllm_config,
self.device)
self.model = self.load_lora_model(self.model, self.vllm_config, self.device)
def get_model(self) -> nn.Module:
return self.model
@@ -90,23 +89,19 @@ class CPUModelRunner(GPUModelRunner):
def _to_list(self, sampled_token_ids: torch.Tensor) -> list[list[int]]:
return sampled_token_ids.tolist()
def get_dp_padding(self,
num_tokens: int) -> tuple[int, Optional[torch.Tensor]]:
def get_dp_padding(self, num_tokens: int) -> tuple[int, Optional[torch.Tensor]]:
# Note: For CPU backend, dp padding is not required for now.
return 0, None
@contextmanager
def _torch_cuda_wrapper():
class _EventPlaceholder:
def __init__(self, *args, **kwargs) -> None:
self.record = lambda: None
self.synchronize = lambda: None
class _StreamPlaceholder:
def __init__(self, *args, **kwargs) -> None:
pass