I submitted a preprocessing docker image a few days ago, which contains only a preprocess.sh. The idea is to perform some preprocessing and then deal with training later. The submission was pending for two days and then I received this error message last night: ``` {"message":"oci runtime error: exec: "/train.sh": stat /train.sh: no such file or directory"} ``` It looks like the system is complaining it cannot find train.sh. Do I have to include train.sh all the time, even for preprocessing only job? It shall be rather easy to modify the system so that it scans for preprocess.sh as well.

Created by Li Shen thefaculty
Hi Li: At the moment if you want 'preprocessing only' you should submit a preprocessing-plus-training submission with a 'dummy' script for training (e.g. an image with a "/train.sh" entry point that does nothing.). We are working on a 'cleaner' method to allow you to do preprocessing only, but it is not ready at this time, and the aforementioned approach will allow you to do what you wish. @josecp: > if you submit a single container ? instead of the two container setting (preprocess + train) ? the system will look for '/train.sh' and try to run that. Even if you're just testing your preprocessing scripts. Yes, that's correct. In particular we would like to avoid inferring your desire to do training, preprocessing, or both by scanning your Docker image. We feel it's safer to ask you to be explicit about your intentions by submitting a text file with the two images specified. > the permissions ? hence access to mount points (/trainingData, /preprocessedData, ...) ? is not the same during training and preprocessing. Yes, again that's correct. We explain that in the section "Training with Preprocessing" here: https://www.synapse.org/#!Synapse:syn4224222/wiki/401759 Thank you.
My experience is that if you submit a single container ? instead of the two container setting (preprocess + train) ? the system will look for '/train.sh' and try to run that. Even if you're just testing your preprocessing scripts. Also, note that the permissions ? hence access to mount points (/trainingData, /preprocessedData, ...) ? is not the same during training and preprocessing. PS- I'm not with the organization of the challenge; I'm just a simple competitor that has bumped into the same issue :)

Does a preprocessing docker image have to include train.sh? page is loading…