diff --git a/modelscope/models/base/base_torch_model.py b/modelscope/models/base/base_torch_model.py
index 2caeb41b..27960b46 100644
--- a/modelscope/models/base/base_torch_model.py
+++ b/modelscope/models/base/base_torch_model.py
@@ -30,10 +30,22 @@ class TorchModel(Model, torch.nn.Module):
 
     def __call__(self, *args, **kwargs) -> Dict[str, Any]:
         # Adapting a model with only one dict arg, and the arg name must be input or inputs
+
+        print(f'\n>>>TorchModel.__call__ args: {args}, kwargs: {kwargs}')
+
         if func_receive_dict_inputs(self.forward):
-            return self.postprocess(self.forward(args[0], **kwargs))
+            # return self.postprocess(self.forward(args[0], **kwargs))
+            res = self.forward(args[0], **kwargs)
+            print(
+                f'>>res in TorchModel.__call__ with func_receive_dict_inputs: {res}'
+            )
+            return self.postprocess(res)
+
         else:
-            return self.postprocess(self.forward(*args, **kwargs))
+            # return self.postprocess(self.forward(*args, **kwargs))
+            res = self.forward(*args, **kwargs)
+            print(f'>>res in TorchModel.__call__: {res}')
+            return self.postprocess(res)
 
     def _load_pretrained(self,
                          net,
diff --git a/modelscope/models/nlp/chatglm2/text_generation.py b/modelscope/models/nlp/chatglm2/text_generation.py
index c9d63724..fa5997b0 100644
--- a/modelscope/models/nlp/chatglm2/text_generation.py
+++ b/modelscope/models/nlp/chatglm2/text_generation.py
@@ -1205,7 +1205,17 @@ class ChatGLM2ForConditionalGeneration(ChatGLMPreTrainedModel):
             **kwargs
         }
         inputs = self.build_inputs(tokenizer, query, history=history)
+
+        print(
+            f'\n>>inputs in ChatGLM2ForConditionalGeneration._chat:\n {inputs}, \n>shape: {inputs.shape}'
+        )
+
         outputs = self.generate(**inputs, **gen_kwargs)
+
+        print(
+            f'\n>>outputs in ChatGLM2ForConditionalGeneration._chat:\n {outputs}, \n>shape: {outputs.shape}'
+        )
+
         outputs = outputs.tolist()[0][len(inputs['input_ids'][0]):]
         response = tokenizer.decode(outputs)
         response = self.process_response(response)
diff --git a/modelscope/pipelines/nlp/llm_pipeline.py b/modelscope/pipelines/nlp/llm_pipeline.py
index d509fcc0..b6b157e1 100644
--- a/modelscope/pipelines/nlp/llm_pipeline.py
+++ b/modelscope/pipelines/nlp/llm_pipeline.py
@@ -121,6 +121,7 @@ class LLMPipeline(Pipeline):
 
         if hasattr(self.model, 'generate'):
             outputs = self.model.generate(**tokens, **forward_params)
+            print(f'>>>self.model.generate: {self.model.generate}')
         elif hasattr(self.model, 'model') and hasattr(self.model.model,
                                                       'generate'):
             outputs = self.model.model.generate(**tokens, **forward_params)