UNAVAILABLE: Invalid argument: unexpected inference input 'sigmoid_0.tmp_0', allowed inputs are: save_infer_model/scale_0.tmp_0

环境

• 【FastDeploy版本】： docker pull registry.baidubce.com/paddlepaddle/fastdeploy:1.0.7-gpu-cuda11.4-trt8.5-21.10
• 【编译命令】如果您是自行编译的FastDeploy，请说明您的编译方式（参数命令）
  Docker内启动服务时：fastdeployserver --model-repository=./examples/vision/ocr/PP-OCR/serving/fastdeploy_serving/models
• 【系统平台】: Linux x64(centos)
• 【硬件】： 说明具体硬件型号，如 Nvidia GPU A10
• 【编译语言】： Python(3.6）

问题日志及出现问题的操作流程

• 【模型跑不通】
• • 先执行examples下的部署示例，可以正确执行,但自己的模型不能运行
    | Model | Version | Status |
    | cls_postprocess | 1 | READY |
    | cls_pp | 1 | READY |
    | cls_runtime | 1 | READY |
    | det_postprocess | 1 | READY |
    | det_preprocess | 1 | READY |
    | det_runtime | 1 | UNAVAILABLE: Invalid argument: unexpected inference input 'sigmoid_0.tmp_0', allowed inputs are: save_infer_model/scale_0.tmp_0 |
    | rec_postprocess | 1 | READY |
    | rec_pp | 1 | READY |
    | rec_runtime | 1 | READY |

Start with the model repository at examples/vision/ocr/PP-OCR/serving/fastdeploy_serving/models and the fastdeployserver command shown in the issue. Compare the det_runtime model's configured input with the reported sigmoid_0.tmp_0 and save_infer_model/scale_0.tmp_0 names. Done means det_runtime reaches READY and the custom OCR model runs through the serving pipeline.