chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

Introduce Stable LoRA ID System for Overlapped Updates and Prefix Caching (#8261)

Browse Source
This commit is contained in:
Lifu Huang
2025-07-23 00:32:16 -07:00
committed by GitHub
parent e885bfdc6a
commit 8abd3e77fe
11 changed files with 400 additions and 261 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/tp_worker.py
+2 -4
View File
@@ -293,11 +293,9 @@ class TpModelWorker:
return parameter
def load_lora_adapter(self, recv_req: LoadLoRAAdapterReqInput):
result = self.model_runner.load_lora_adapter(
recv_req.lora_name, recv_req.lora_path
)
result = self.model_runner.load_lora_adapter(recv_req.to_ref())
return result
def unload_lora_adapter(self, recv_req: UnloadLoRAAdapterReqInput):
result = self.model_runner.unload_lora_adapter(recv_req.lora_name)
result = self.model_runner.unload_lora_adapter(recv_req.to_ref())
return result