Failed to start server error when trying run docker using the tensorflow/serving image

Question

I am using the tensorflow/serving image and run into this error trying run the docker file.

Step 1: > pulled the tensorflow serving image

docker pull tensorflow/pull

Step2: Stored my models into a (saved-models) folder within my project folder(detection-potato-lite)

Elijah-A-W@DESKTOP-34M2E8U MINGW64 /d/myn/ML_Prediction_Project/New_folder/detection-potato-lite
$ ls
   Api/  'Data Eda'/   models.config   Plant/   saved-models/

Step 3: Created the models.config file in the Project folder to render the models dynamically using tensorflow serving

    model_config_list {
    config {
        name: 'potatoes_model'
        base_path: '/detection-potato-lite/saved-models'
        model_platform: 'tensorflow'
        model_version_policy: {all: {}}
}

Step4: Get an error(No such file/directory) when trying run the image in Windows Powershell

PS C:UsersElijah-A-W> docker run -t --rm -p 8501:8501 -v D:/myn/ML_Prediction_Project/New_Folder/detection-potato-lite:/detection-potato-disease tensorflow/serving --rest_api_port=8501 --model_config_file=/detection-potato-lite/models.config

Failed to start server. Error: Not found: /detection-potato-lite/models.config; No such file or directory

Asked By: andy lacron

||

Source

Answer 1

I see two problems here:

The directory path you bind to the volume is different to the one you list locally (New_Folder vs New_folder)
You bind local directory ‘detection-potato-lite’ to docker directory ‘detection-potato-disease’. When you specify the location of the config file and base-path for your application to run in docker, it must be relative to where you store it in docker so it will be ‘detection-potato-disease’.

EDIT:

By using -v D:/myn/ML_Prediction_Project/New_Folder/detection-potato-lite:/detection-potato-disease in your docker run command, you are binding your local directory (D:/myn/ML_Prediction_Project/New_Folder/detection-potato-lite) to a directory within the Docker container (/detection-potato-disease).

This means that when your application runs inside the Docker container, all your files/folders that were on that local directory now exist in ‘/detection-potato-disease’.

So you have two options:

Modify the ‘base_path’ in your models.config file from /detection-potato-lite/saved-models to /detection-potato-disease/saved-models AND modify ‘–model_config_file’ in your docker run command from --model_config_file=/detection-potato-lite/models.config to --model_config_file=/detection-potato-disease/models.config
Change your docker run command to be docker run -t --rm -p 8501:8501 -v D:/myn/ML_Prediction_Project/New_folder/detection-potato-lite:/detection-potato-lite tensorflow/serving --rest_api_port=8501 --model_config_file=/detection-potato-lite/models.config

EDIT 2:

You’ll need to add another closing } bracket on line 7 of the models.config file after the previous fix I have mentioned.

Answered By: JasonS

Failed to start server error when trying run docker using the tensorflow/serving image

Question:

Answers: