A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Bugfix] Fix GPUModelRunner has no attribute lora_manager (#24762)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Jee Jee Li committed
Sep 13, 2025 at 14:55 UTC
99bfef841fc705c17a46b364b324da86a2aee3f9
1 file changed
+10
-10
vllm/v1/worker/lora_model_runner_mixin.py
+10
-10
@@ -63,8 +63,7 @@ class LoRAModelRunnerMixin:
63
def _set_active_loras(self, prompt_lora_mapping: tuple[int, ...],
64
token_lora_mapping: tuple[int, ...],
65
lora_requests: set[LoRARequest]) -> None:
66
- if not self.lora_manager:
67
- raise RuntimeError("LoRA is not enabled.")
66
+ self._ensure_lora_enabled()
67
68
# Set is_prefill to True, so we always use the SGMV kernels on
69
# non-cuda platforms.
@@ -75,6 +74,11 @@ class LoRAModelRunnerMixin:
74
is_prefill=True)
75
self.lora_manager.set_active_adapters(lora_requests, lora_mapping)
76
77
+ def _ensure_lora_enabled(self) -> None:
78
+ if not hasattr(self, "lora_manager"):
79
+ raise RuntimeError(
80
+ "LoRA is not enabled. Use --enable-lora to enable LoRA.")
81
+
82
def set_active_loras(self, input_batch: InputBatch,
83
num_scheduled_tokens: np.ndarray) -> None:
84
@@ -172,21 +176,17 @@ class LoRAModelRunnerMixin:
176
self.lora_manager.remove_all_adapters()
177
178
def add_lora(self, lora_request: LoRARequest) -> bool:
175
- if not self.lora_manager:
176
- raise RuntimeError("LoRA is not enabled.")
179
+ self._ensure_lora_enabled()
180
return self.lora_manager.add_adapter(lora_request)
181
182
def remove_lora(self, lora_id: int) -> bool:
180
- if not self.lora_manager:
181
- raise RuntimeError("LoRA is not enabled.")
183
+ self._ensure_lora_enabled()
184
return self.lora_manager.remove_adapter(lora_id)
185
186
def pin_lora(self, lora_id: int) -> bool:
185
- if not self.lora_manager:
186
- raise RuntimeError("LoRA is not enabled.")
187
+ self._ensure_lora_enabled()
188
return self.lora_manager.pin_adapter(lora_id)
189
190
def list_loras(self) -> set[int]:
190
- if not self.lora_manager:
191
- raise RuntimeError("LoRA is not enabled.")
191
+ self._ensure_lora_enabled()
192
return self.lora_manager.list_adapters()