platipy icon indicating copy to clipboard operation
platipy copied to clipboard

urllib error on clusters in run_hybrid_segmentation

Open Ken32g opened this issue 11 months ago • 2 comments

When running run_hybrid_segmentation on HPC clusters, it may report the following error: urllib.error.URLError: <urlopen error [Errno -2] Name or service not known>. This error is likely raised because the cluster nodes cannot access the data URL. Therefore, run_hybrid_segmentation needs to be executed on a login node for the first time. After the dataset has been downloaded, the task can be submitted onto the HPC cluster, and it should work properly.

Ken32g avatar Mar 26 '24 05:03 Ken32g

Hi @Ken32g, thanks for flagging this. The tool will download the models etc on first run so I this wouldn't work if there is no internet connection on the cluster. It sounds like you have a solution but in case it is useful, check the .platipy directory in your users home directory as this is where it downloads the models. You can copy that directory to other locations to bypass having to redownload the model everytime.

pchlap avatar Mar 26 '24 14:03 pchlap

Thank you!

Ken32g avatar Mar 26 '24 16:03 Ken32g