Open ant0nsc opened 3 years ago
ant0nsc
commented
3 years ago Example of how that could be invoked:
python download_files.py --run abc_123 --pattern segmentation.nii.gz would download all files that contain "segmentation.nii.gz"--pattern "Test/*/segmentation.nii.gz" would restrict to the test folder. pattern effectively a regex?
kh296
commented
3 years ago The syntax looks fine to me, and making pattern effectively a regex sounds like a good idea. It would also be nice to be able to specify a list of patterns, for example:
--patern "Test/*/segmentation.nii.gz Test/metrics.csv"
The folder hierarchy after download should probably be the same as before download. For example:
--run abc_123 --patern "Test/*/segmentation.nii.gz" would download to
abc_123/outputs/epoch*/Test/*/segmentation.nii.gz;
--run abc_123 --patern "Test/metrics.csv" would download to abc_123/outputs/CrossValResults/*/epoch*/Test/metrics.csv .
Allow user to specify subset of results for easy download => At present I can navigate to an experiment’s azureml/ExperimentRun/dcid.my_run_id/outputs folder, but then the options for bulk download are a bit constrained. As a specific use case, I’d like to be able to download segmentation.nii.gz for all patients in the test dataset. These are all under outputs/epoch_xyz/Test, but together with the individual-organ segmentations. For head-and-neck patients, being able to download only the segmentation.nii.gz would reduce the data volume by about a factor of 20.
@kh296
AB#4218