Why docker run command doesn't work

[hanhanwu]

I'm following every step in the quick start: https://colab.research.google.com/github/bentoml/BentoML/blob/master/guides/quick-start/bentoml-quick-start-guide.ipynb

Also new to docker. I'm using Windows 10. Want to see how to run the built image through docker.
Every step till bentoml containerize IrisClassifier:latest -t iris-classifier works fine in my local ipython notebook.
A folder had been built in my local "\bentoml\repository\IrisClassifier\20200929232341_4535FA".

But I tried to run command docker run -p 5000:5000 iris-classifier:latest --workers=1 --enable-microbatch through ipython or powershell, often got this error:

Do you know how can I run this docker command successfully?

[yubozhao]

@hanhanwu Thank you for point it out. We missed the :latest on the bentoml containerize command. It should be:
bentoml containerize IrisClassifier:latest -t iris-classifier:latest

In your case, the docker image is properly built. You can use docker images command to list all of the docker images on your system, and you should see it in there. The docker image tag will be your IrisClassifier version 20200928232341_4535FA.

If you see the image, you can run docker run -p 5000:5000 iris-classifier:20200928232341_4535FA --workers=1 --enable-microbatch to test it out.

Right now, the bentoml containerize behavior is if you provide a docker image tag without :, it will default to use the bento's version instead. For this case, with --tag iris-classifier, the finished docker tag will be iris-classifier:BENTO_VERSION

[hanhanwu]

Great! Thank you!
Now I'm able to see the image running in the docker!

One more question, then how can I use this running image? Can I use it the same as REST API to give it some input and get the output? Or how can I run the trained model through docker?

[yubozhao]

@hanhanwu Awesome, glad you have it running

Yes, you can use the running docker container as REST API server. If you have curl installed on your system. You can run the same curl command.

curl -i
--header "Content-Type: application/json"
--request POST
--data '[[5.1, 3.5, 1.4, 0.2]]'
localhost:5000/predict
You can also visit the index page http://localhost:5000 use the swagger UI.

how can I use this running image

 Once you build a docker image, you can deploy to a lot of places such as Kubernetes, AWS ECS and etc.  
 For your running image (container), it only work locally to your system, unless you have open your system for 
 others to access over network.

how can i run the trained model through docker

 I am not so sure I understand your question, can you elaborate? 

[hanhanwu]

Perfect, this did answered my question. Thank you very much @yubozhao !