![cellprofiler analyst example properties file cellprofiler analyst example properties file](https://global.discourse-cdn.com/business4/uploads/imagej/original/3X/6/b/6b1539826e153afdd3760e438ba76996d9483c76.png)
Once you run all the cells in sequence, you should be able to view your results with CellProfiler Analyst by opening the. Combine all object measurements in one.Modify image file paths (from Ukko2 to the local environment).Copy the results from Ukko2 to the local computer.The notebook "cp_on_ukko2.ipynb" contains further cells that implement the following tasks: Use paramiko to make SSH connention and run the script.Allow SSH connections using the key previously created to run the script.Create a script that finds and runs the newest sbatch script in a directory (also check that output files don't exist yet).If this step proves too cumbersome for users, it should be possible to make it work from the notebook with the following steps: Paste the command in the terminal and hit Enter. Once you have the Ukko2 terminal open, go back to the notebook and copy the command it printed out. Type "ssh ukko2.cs.helsinki.fi", press Enter.The script also prints out the command you need to run on Ukko2 to submit the job. This should create a folder with the batch job script in Ukko2 $PROJ. These will be used to name the output folders. Set your email address (in the hope that email notifications start to work, at the moment they don't).Change 'user' to your university username.Now you have to modify the script to your situation: Start the notebook "cp_on_ukko2.ipynb".Download LMU example notebooks from (download all files in one folder).Set up Jupyter Notebook (see Jupyter Notebook quick start).Learning to work with Python notebooks can also be useful for analysing the results. To make things a bit easier, you can use a Python notebook provided by LMU to create your batch job scripts. You need to create a batch job script that describes how to run your analysis.
![cellprofiler analyst example properties file cellprofiler analyst example properties file](https://i.ytimg.com/vi/4H6VqjnWpUw/maxresdefault.jpg)
The jobs are started on Linux command line.Starting compute jobs on a cluster is a bit different from running it on your own desktop (see FAQ & Scientific Software Use Cases): ExportToDatabase using SQLite does not work on the cluster.Use "Default Output Folder" in all modules that write results.For this to work, you need to provide a mapping between the image data paths on your local machine and on the cluster (see screenshot). When the pipeline is ready and you click "Analyze Images", instead of actually running the analysis, CreateBatchFiles will create a file that contains all necessary information for running the pipeline on the cluster. The main difference is that you need to add "CreateBatchFiles" as the last module. You can still build and test your CP pipeline in the usual way with the GUI. Here's how to access the folders on your local computer. In $PROJ you can store the scripts needed to start the computation, and also the results. You need to copy your image data in the $WRKDIR directory. There are two important folders on Ukko2: $WRKDIR and $PROJ (see OBSOLETE - Ukko2 User Guide). Mount Ukko2 folders on the local computer A speedup like this could be quite significant: analyses that you used to let run over night could now be repeated several times times during the day with different parameters. The benefit of Ukko2 was that you would get 8 times more images processed simultaneously. In this case the time it took to process one image set was about the same on Ukko2 and HCA workstation. The pipeline was reasonably heavy, with texture measurements etc., and the run would take 10-15 minutes. For a test data set of 144 images sets the processing used 12 cores on each of the 4 nodes (total of 48 cores), with each core processing 3 image sets. How many of those will actually be used depends also on the data set size. At the moment there is some problem when trying to use all cores, but using even just half still gives 56 cores (CP workers). One node has 28 cores, giving 112 cores in total. On Ukko2, a modest resource request that based on the testing done so far is usually granted immediately is 4 nodes (sometimes the job would be in queue for maybe 5 minutes). When analysing Molecular Devices Nano data (2048x2048) on LMU HCA workstation you can run only 6 workers before getting memory problems. To get access, see: OBSOLETE - Ukko2 User Guide.You need to learn a few things to get it working, but the time investment may pay off if you have lots of plates to process. This page shows how to run CellProfiler on Computer Science department's Ukko2 cluster ( OBSOLETE - Ukko2 User Guide).