Update app.py

app.py

@@ -134,7 +134,7 @@ def main():
 
     if ptuning_checkpoint is not None:
         print(f"Loading prefix_encoder weight from {ptuning_checkpoint}")
-        model = AutoModel.from_pretrained("THUDM/chatglm-6b-int4", config=config, trust_remote_code=True).float()
+        model = AutoModel.from_pretrained("THUDM/chatglm-6b-int4", config=config, trust_remote_code=True)
         prefix_state_dict = torch.load(os.path.join(ptuning_checkpoint, "pytorch_model.bin"), map_location='cpu')
         new_prefix_state_dict = {}
         for k, v in prefix_state_dict.items():
@@ -142,12 +142,12 @@ def main():
                 new_prefix_state_dict[k[len("transformer.prefix_encoder."):]] = v
         model.transformer.prefix_encoder.load_state_dict(new_prefix_state_dict)
     else:
-        model = AutoModel.from_pretrained("THUDM/chatglm-6b-int4", config=config, trust_remote_code=True).float()
+        model = AutoModel.from_pretrained("THUDM/chatglm-6b-int4", config=config, trust_remote_code=True)
 
     model = model.quantize(4)
 
     # P-tuning v2
-    model = model.half()
+    # model = model.half()
     model.transformer.prefix_encoder.float()
     
     model = model.eval()