Update handler.py

handler.py

@@ -1,5 +1,5 @@
 # https://github.com/sayakpaul/diffusers-torchao
-#8s
+#6.22s
 import os
 from typing import Any, Dict
 from PIL import Image
@@ -18,8 +18,8 @@ class EndpointHandler:
             torch_dtype=torch.bfloat16,
         ).to("cuda")
         apply_cache_on_pipe(self.pipe, residual_diff_threshold=0.12)
-        # quantize_(self.pipe.text_encoder, float8_weight_only())
-        # quantize_(self.pipe.transformer, float8_dynamic_activation_float8_weight())
+        quantize_(self.pipe.text_encoder, float8_weight_only())
+        quantize_(self.pipe.transformer, float8_dynamic_activation_float8_weight())
         self.pipe.transformer = torch.compile(
             self.pipe.transformer, mode="max-autotune-no-cudagraphs",
         )