---
title: "辨識錯誤 tensorflow KeyError \"The name 'import/input' refers to an Operation not in the graph.\""
description: "簡單來解釋就是他的名字你打錯了 所以 input 才會 input 錯"
author: "kidneyweakx"
published: 2019-04-07T00:00:00.000Z
updated: 2026-06-02T17:22:05.558Z
lang: zh-TW
tags: ["程式筆記", "Tensorflow"]
canonical: https://kidneyweakx.com/blog/19log-0407-tensor-error
---

# 辨識錯誤 tensorflow KeyError "The name 'import/input' refers to an Operation not in the graph."

### **近期遇到的小錯誤**

```python
KeyError: "The name 'import/input' refers to an Operation not in the graph."
```

![1.00](https://raw.githubusercontent.com/kidneyweakx/img-host/image/image/tensor01.PNG)

簡單來解釋就是他的名字你打錯了 所以 input 才會 input 錯

我自己爬文去嘗試了各種` "Mul" "input"`之類的都錯

最後看到這篇文章 [link](https://developer.arm.com/technologies/machine-learning-on-arm/developer-material/how-to-guides/optimizing-neural-networks-for-mobile-and-embedded-devices-with-tensorflow/determine-the-names-of-input-and-output-nodes)

### **懶人包**

> 先用 `tensorflow` 的工具將 `.pb` 轉成可以用 `tensorboard` 可以觀看的格式，再查出 `input_layer`(我的是 `placeholder`)和 `output_layer` 是甚麼名字再重新修正。

這是我的程式碼(再依照個人需求更改)

```python
python tensorflow\examples\label_image\label_image.py --graph=tmp\output_graph.pb --label=tmp\output_labels.txt --image=tmp\91076.png --input_layer=Placeholder --output_layer=final_result
```

輸出
![](https://raw.githubusercontent.com/kidneyweakx/img-host/image/image/tensor02.PNG)