xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 03:35:17 +08:00
Code Issues Packages Projects Releases Wiki Activity

Improve static type checking in LoRAModelRunnerMixin (#17104)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor 2025-04-24 14:14:47 +01:00 committed by GitHub
parent b0c1f6202d
commit 21f4f1c9a4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 6 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
vllm/v1/worker/lora_model_runner_mixin.py
View File

@ -28,20 +28,16 @@ class LoRAModelRunnerMixin:
scheduler_config: SchedulerConfig,
lora_config: LoRAConfig, device: str) -> nn.Module:
assert supports_lora(
model), f"{model.__class__.__name__} does not support LoRA yet."
if not supports_lora(model):
raise ValueError(
f"{model.__class__.__name__} does not support LoRA yet.")
if supports_multimodal(model):
logger.warning("Regarding multimodal models, vLLM currently "
"only supports adding LoRA to language model.")
# It's necessary to distinguish between the max_position_embeddings
# of VLMs and LLMs.
if hasattr(model.config, "max_position_embeddings"):
max_pos_embeddings = model.config.max_position_embeddings
else:
max_pos_embeddings = (
model.config.text_config.max_position_embeddings)
# Use get_text_config() in case of multimodal models
text_config = model_config.hf_config.get_text_config()
# Add LoRA Manager to the Model Runner
self.lora_manager = LRUCacheWorkerLoRAManager(
@ -52,7 +48,7 @@ class LoRAModelRunnerMixin:
device,
model.embedding_modules,
model.embedding_padding_modules,
max_position_embeddings=max_pos_embeddings,
max_position_embeddings=text_config.max_position_embeddings,
)
return self.lora_manager.create_lora_manager(model)