通过在单独的 Z05B6053C41A2130AFDZ6FCDAE1 中运行的 R 实例在本地保存 R model

Question

Hi I am currently building some models using R connected to a remote h2o instance running in a docker container.

I am at a point where I would like to iterate through a leaderboard of models and h2o.saveModel() each model to my local computer.

However, since saveModel() saves to the computer running h2o, my models are being saved within the docker container.

The solution I have come up with is to scp all the model files to my computer after h2o has finished running them, however I was wondering if anyone out there had run into this problem and had come up with a better solution.

Note: Running a local h2o instance for building the models is not an option

Thank you for the help

Answer 1

This implementation is a good use case to leverage docker volumes.

Create an external volume and mount/bind it to the location where the model is being saved.

https://docs.docker.com/storage/volumes/

Answer 2

For anyone that might run into this problem in the future, I used the R ssh library ( https://www.rdocumentation.org/packages/ssh/versions/0.6 ) to solve the problem.

After saving all my models I run a docker cp command to copy the files from the docker container to the host running the docker.

Then I used the scp_download function from the ssh library to download the files to my computer.

Finally, I ran 2 ssh_exec_wait functions to delete the folders created on the the docker host as well as in the docker container itself.