remove handler

.gitignore

@@ -1,3 +1,4 @@
 test_handler.py
 audio1.wav
-test_online_inference.py
+test_online_inference.py
+handler_tmp.py

handler.py

@@ -42,7 +42,8 @@ class EndpointHandler():
         """
 
         inputs = data.pop("inputs", data)
-        with torch.cuda.amp.autocast():
+        # with torch.cuda.amp.autocast():
+        with torch.no_grad():
             prediction = self.pipeline(inputs, generate_kwargs={"forced_decoder_ids": self.forced_decoder_ids}, max_new_tokens=255)
             # prediction = self.pipeline(inputs, return_timestamps=False)
             prediction['text'] = prediction['text'] + '????'