Upload handler.py

handler.py

@@ -10,14 +10,15 @@ torch._dynamo.config.suppress_errors = True
 
 #from huggingface_inference_toolkit.logging import logger
 
-def compile_pipeline(pipe):
+def compile_pipeline(pipe) -> Any:
+    pipe.transformer.fuse_qkv_projections()
     pipe.transformer.to(memory_format=torch.channels_last)
     pipe.transformer = torch.compile(pipe.transformer, mode="reduce-overhead", fullgraph=False, dynamic=False, backend="inductor")
     return pipe
 
 class EndpointHandler:
     def __init__(self, path="", **kwargs: Any) -> None:  # type: ignore
-        is_compile = False
+        is_compile = True
         #repo_id = "camenduru/FLUX.1-dev-diffusers"
         repo_id = "NoMoreCopyright/FLUX.1-dev-test"
         dtype = torch.bfloat16