Hi, We would like to add ~2 GB of preprocessed date to the preprocessed folder. However, we would like to keep the current data in the preprocessed folder as well. Rerunning the whole preprocessing stage will consume 3 days of server time. Is there a way we can modify / add file to the preprocessed folder without losing the current data? Alternatively, is there a size limit on the uploaded image? We estimate the compressed preprocessed data will be ~1GB. Will it possible to temporarily extend our modelstate folder to 2-3Gb, so we can drop the additional prepossessed information to the modelState folder, and the upload it with every image? Thanks ahead, Eli

Created by Eli Meirom bloodymeli
> Is there a way we can modify / add file to the preprocessed folder without losing the current data? I'm sorry but we do not support that. > Alternatively, is there a size limit on the uploaded image? Yes, your Docker installation by default imposes a 10GB limit on model size. We ask that you stay within this limit and reserve the right to reject images that exceed this size. > Will it possible to temporarily extend our modelstate folder to 2-3Gb, so we can drop the additional prepossessed information to the modelState folder, and the upload it with every image? The /modelState folder is not a mechanism for uploading data files but rather a mechanism for retrieving the results of model training that takes place on the challenge servers.   To summarize: If you have 1GB of data that you wish to use along with your existing preprocessing data while training your model, you may include the data in your submitted Docker training image. As a reminder, to preserve the cached preprocessing results you must specify in your new submission the same preprocessing image that was used to produce the cached output. The system will then recognize that the preprocessing was already done and advance to the (new) training image.

Modifying preProcessed folder files page is loading…