vllm.lora.request ¶
LoRARequest ¶
Bases: Struct
Request for a LoRA adapter.
lora_int_id must be globally unique for a given adapter. This is currently not enforced in vLLM.
Source code in vllm/lora/request.py
base_model_name class-attribute instance-attribute ¶
base_model_name: str | None = field(default=None)
lora_local_path class-attribute instance-attribute ¶
lora_local_path: str | None = field(default=None)
tensorizer_config_dict class-attribute instance-attribute ¶
tensorizer_config_dict: dict | None = None
__eq__ ¶
Overrides the equality method to compare LoRARequest instances based on lora_name. This allows for identification and comparison lora adapter across engines.
Source code in vllm/lora/request.py
__hash__ ¶
__hash__() -> int
Overrides the hash method to hash LoRARequest instances based on lora_name. This ensures that LoRARequest instances can be used in hash-based collections such as sets and dictionaries, identified by their names across engines.