Add env var to control custom Triton kernel cache and set CSGMV as default backend. (#12176)

This commit is contained in:
Lifu Huang
2025-10-27 17:49:32 -07:00
committed by GitHub
parent 88596739a4
commit ce832d7034
13 changed files with 24 additions and 25 deletions
+1 -1
View File
@@ -764,7 +764,7 @@ class LoRAUpdateTestSessionBase:
max_lora_rank: Optional[int],
enable_lora: Optional[bool] = None,
lora_target_modules: Optional[List[str]] = None,
lora_backend: str = "triton",
lora_backend: str = "csgmv",
disable_cuda_graph: bool = False,
cuda_graph_max_bs: int = 4,
):