update model_worker.py

model_worker.py

@@ -232,7 +232,9 @@ class ModelWorker:
         except ImportError:
 
             def install_flash_attn():
-                os.system("pip install flash-attn==2.5.9.post1")
+                os.system(
+                    "FLASH_ATTENTION_SKIP_CUDA_BUILD=TRUE pip install flash-attn==2.5.9.post1 --no-build-isolation"
+                )
 
             install_flash_attn()
             # import flash_attn