Update app.py

app.py

@@ -17,22 +17,27 @@ pipeline_optimized = FluxPipeline.from_pretrained(
     "camenduru/FLUX.1-dev-diffusers",
     torch_dtype=torch.bfloat16
 ).to("cuda")
-pipeline_optimized.transformer = autoquant(
-    pipeline_optimized.transformer
-)
-print("all good")
 pipeline_optimized.transformer.to(memory_format=torch.channels_last)
 pipeline_optimized.transformer = torch.compile(
     pipeline_optimized.transformer,
     mode="max-autotune",
     fullgraph=True
 )
+# wrap the autoquant call in a try-except block to handle unsupported layers
+for name, layer in pipeline_optimized.transformer.named_children():
+    try:
+        # apply autoquant to each layer
+        pipeline_optimized.transformer._modules[name] = autoquant(layer, error_on_unseen=False)
+        print(f"Successfully quantized {name}")
+    except AttributeError as e:
+        print(f"Skipping layer {name} due to error: {e}")
+    except Exception as e:
+        print(f"Unexpected error while quantizing {name}: {e}")
 
-
-pipeline_optimized.transformer = autoquant(
-    pipeline_optimized.transformer,
-    error_on_unseen=False
-)
+# pipeline_optimized.transformer = autoquant(
+#     pipeline_optimized.transformer,
+#     error_on_unseen=False
+# )
 pipeline_normal = pipeline_optimized
 
 @spaces.GPU(duration=120)