onnxruntime输出模型中间层的结果

        def get_layer_output(model, image):
            ori_output = copy.deepcopy(model.graph.output)
            for node in model.graph.node:
                for output in node.output:
                    model.graph.output.extend([onnx.ValueInfoProto(name=output)])
            ort_session = rt.InferenceSession(model.SerializeToString(),providers=["CPUExecutionProvider"])
            ort_inputs = {}
            for i, input_ele in enumerate(ort_session.get_inputs()):
                ort_inputs[input_ele.name] = image
            outputs = [x.name for x in ort_session.get_outputs()]
            ort_outs = ort_session.run(outputs, ort_inputs)
            print(type(ort_outs))
            print(type(ort_outs[0]))
            # print("ort_outs[0]",ort_outs[0])
            # exit()    
            # ort_outs = OrderedDict(zip(outputs, ort_outs))
            return ort_outs     
                 
        model_path = "/data_alpha/test_1221/MobileSAM-pytorch/exp/adamw_lr_1e-3_wd_5e-4_bs_8_epoch_16/ckpt/tiny_2262.onnx"
        onnx_model = onnx.load(model_path)
                 
        # print(outputs)        
        intermediate_layer_names = [onnx_model.graph.node[i].name for i in range(len(onnx_model.graph.node))]
        # print(intermediate_layer_names)
        for layer_name, output_data in zip(intermediate_layer_names, get_layer_output(onnx_model,tensor_input_2)):
            if "Conv_" not in layer_name:
                continue        
            # output_file_path = f"/tmp/onnx_txt/{layer_name}.txt"
            # output_file_path = f"{output_folder}/{layer_name}.txt"
            # np.savetxt(output_file_path, output_data.flatten().astype(np.float32), fmt='%f')
        # exit()