从 bash 运行 tensorflow_model_server 会引发“找不到命令错误”答案

【问题标题】：Running tensorflow_model_server from bash is throwing 'command not found error'从 bash 运行 tensorflow_model_server 会引发“找不到命令错误”
【发布时间】：2018-11-23 08:58:22
【问题描述】：

我正在从 bash 运行 tensorflow 服务器：使用的命令：

tensorflow_model_server --rest_api_port=8501 --model_name=saved_model --model_base_path=$(<path to .pb file in my system>/saved_model.pb)

我收到以下错误：

<path to .pb file>/saved_model.pb:
 line 1: $'\b\001\022\344\307\376\016': command not found <path to .pb file>/saved_model.pb:
 line 2: $'\203\020\022\345\017': command not found <path to .pb file>/saved_model.pb:
 line 4: $'\003Add\022\006': command not found <path to .pb file>/saved_model.pb:
 line 6: $'\001x\001T\022\006\n\001y\001T\032\006': command not found <path to .pb file>/saved_model.pb:
 line 7: $'\001z\001T\E': command not found <path to .pb file>/saved_model.pb:
 line 8: $'\001T\022\004type:\020': command not found <path to .pb file>/saved_model.pb:
 line 9: $'\0162\f\016\023\001\002\004\006\005\003': command not found <path to .pb file>/saved_model.pb:
 line 10: 8: command not found <path to .pb file>/saved_model.pb:
 line 11: $'\004Cast\022': command not found <path to .pb file>/saved_model.pb:
 line 14: $'\001x\004SrcT\032\t\n\001y\004DstT\f\n\004SrcT\022\004type\f': command not found <path to .pb file>/saved_model.pb:
 line 15: $'\004DstT\022\004type': command not found <path to .pb file>/saved_model.pb:
 line 16: 8: command not found <path to .pb file>/saved_model.pb:
 line 17: $'\005Const\032\017': command not found <path to .pb file>/saved_model.pb:
 line 18: $'\006output\005dtype\017': command not found <path to .pb file>/saved_model.pb:
 line 24: $'\005value\022\006tensor\r\n\005dtype\022\004type\n\354\001\n\006Conv2D\022\n\n\005input\001T\022\v': command not found <path to .pb file>/saved_model.pb:
 line 28: $'\006filter\001T\032\v\n\006output\001T\023\n\001T\022\004type:\b\n\0062\004\023\016\001\002\024': command not found <path to .pb file>/saved_model.pb:
 line 29: syntax error near unexpected token (' <path to .pb file>/saved_model.pb:
 line 29: strides�  list(int)"�'

我不确定如何解决此问题。我在我的系统中使用 anaconda。任何线索都会非常有帮助。

提前致谢

【问题讨论】：

标签： python bash tensorflow protocol-buffers tensorflow-serving

【解决方案1】：

model_base_path应该是目录，是saved_model.pb和variables\的版本文件夹的父文件夹，不是文件。例如，对于以下目录，model_base_path 应为 path_to_versions\versions。

`-- versions
    `-- 1
        |-- saved_model.pbtxt
        `-- variables
            |-- variables.data-00000-of-00001
            `-- variables.index

【讨论】：

感谢@Yiding。我相应地进行了更改，“model_base_path”现在指向包含 .pb 文件和变量文件夹的目录。但是，这种尝试也引发了错误。使用的命令（/serve/1 文件夹包含 .pb 文件）： tensorflow_model_server --rest_api_port=8501 --model_name=saved_model --model_base_path=$(/CycleGAN-TensorFlow_v2/serve/ 1/) 请在下面找到新错误：bash: /CycleGAN-TensorFlow_v2/serve/1/: Is a directory Illegal instruction