biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V0 deprecation] Remove long context LoRA (#21169)

Browse Source 
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
Jee Jee Li
2025-07-19 17:15:41 +08:00
committed by GitHub
parent cf8cc32674
commit 1eaff27815
13 changed files with 35 additions and 301 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/lora/peft_helper.py
View File

@@ -35,12 +35,9 @@ class PEFTHelper:
use_rslora: bool = field(default=False)
# True to use Weight-Decomposed Low-Rank Adaptation (DoRA, see: https://arxiv.org/abs/2402.09353)
use_dora: bool = field(default=False)
# long context lora field
context_length: int = field(default=0)
# Extra vllm field, start with 'vllm_' to avoid conflict
vllm_lora_scaling_factor: float = field(default=1.0)
vllm_max_position_embeddings: Optional[int] = field(default=False)
vllm_long_context_scaling_factor: Optional[float] = field(default=None)
def _validate_features(self) -> list[str]:
"""
@@ -59,12 +56,6 @@ class PEFTHelper:
self.vllm_lora_scaling_factor = self.lora_alpha / math.sqrt(self.r)
else:
self.vllm_lora_scaling_factor = self.lora_alpha / self.r
if self.context_length:
if self.vllm_max_position_embeddings is None:
self.vllm_max_position_embeddings = self.context_length
self.vllm_long_context_scaling_factor = float(
math.ceil(self.context_length /
self.vllm_max_position_embeddings))
@classmethod
def from_dict(cls, config_dict: dict) -> "PEFTHelper":