SAS Support Communities

RadhikhaMyneni · ‎09-15-2022

Hi Filbert, You run the javac command from the directory that contains the src directory. So in the example it is C:\SGF2015\OpenSrcIntegration. To run as administrator, don't double-click and open the file with default editor but see if you can right-click on the file (from file explorer) and select Run as Administrator. If that does not work, try right-clicking on windows command prompt and Run as Administrator and then go to the directory where the file is and then open using notepad. The sasv9.cfg is a text file so once you have access, you should be able to edit it with notepad. Radhikha

RadhikhaMyneni · ‎09-14-2022

Hi Filbert, The java file(s) (SASJavaExec.java) should be in src/dev directory and when you compile using "javac src/dev/* -d bin" command, the executable will go into bin directory. For your second question, since it is a personal desktop you probably can open the sasv9.cfg file as a administrator - on Windows you typically do this by right-clicking on the file and selecting Run as Administrator. Hope this helps, Radhikha

RadhikhaMyneni · ‎08-25-2022

Here is a blog post on building and deploying a machine learning pipeline using SAS and Python and it talks about registering a Python model that is built in the Open Source Code node (in Visual Machine Learning) to Model Manager.

RadhikhaMyneni · ‎06-14-2021

Hi Andreas, For the plot to show up in the node results, you need to: (1) save it in jpg, png or gif format, (2) name it with rpt_ prefix, for example rpt_treeplot.png or rpt_treeplot.jpg and (3) save it in dm_nodedir folder (dm_nodedir is a variable available in the node editor pointing to a temporary working folder) Here is some sample Python code that does this: # Plot model residuals plt.scatter(pred, pred - dm_inputdf[dm_dec_target]) plt.axhline(y=0, color='r') plt.title('Residual plot') plt.ylabel('Residual') plt.savefig(dm_nodedir + '/rpt_residuals.png') plt.close() Hope this helps, Radhikha

RadhikhaMyneni · ‎03-26-2021

When an error occurs in the Open Source Code node, the generic message Encountered error code 1 when executing PYTHON program is highlighted in the log. The detailed error messages that help pinpoint the problem i.e., the messages coming from Python itself are displayed above this generic message; you can scroll up to view them. Additional debugging questions are documented here: https://go.documentation.sas.com/?cdcId=vdmmlcdc&cdcVersion=8.3&docsetId=vdmmlref&docsetTarget=n09n0yjpv48gddn0z5gv364f95qv.htm&locale=en

RadhikhaMyneni · ‎11-17-2020

Hi, When you say it failed, can you tell if the failure was in Python or afterwards in the Open Source Code node?. Did you place the node in the Supervised Lane? Also, can you post back the failure messages from the log - When an error occurs in the Open Source Code node, the generic message Encountered error code 1 when executing Python program is highlighted in the log. The detailed error messages that help pinpoint the problem are displayed above this generic message and you can scroll up to view them. You can also search for the first occurrence of executeProcess string in the log to see the start of these detailed error messages. I would think your use case should work as the only two columns needed in dm_scoreddf dataframe that the node expects you to create (if it is in Supervised Lane) are posterior probabilities. I am assuming here that you have a binary or nominal target. It should not matter that you label encoded the target and used it in the Python code. Radhikha

RadhikhaMyneni · ‎08-27-2020

Currently the Open Source Code node in Model Studio does not have a way to invoke a specific virtual environment in the Python install but I do agree that it is useful and will try to look into adding in future release. When the node executes, the Python executable configured is used to make a call similar to below code where <fileToRun> is constructed on the fly mostly using code from the node editor. python <fileToRun>

RadhikhaMyneni · ‎04-13-2020

Hi Prajna, You can always turn off data partitioning at project creation time. When creating a project, click on the Advanced button on the "New Project" window, select "Partition Data" tab and un-select "Create partition variable" check-box (see pic below). Note that you will not be able to change this setting after the Data node in any one of the pipelines is run. Radhikha

RadhikhaMyneni · ‎03-31-2020

Hello, Currently there is no way to invoke a python virtual environment from the Open Source Code (OSC) node because the node just calls the python executable it finds (mostly in PATH) and invokes its base environment. You can update the PATH by modifying the sas-compsrv file under the /opt/sas/viya/config/etc/sysconfig/compsrv/default directory by adding the following line: export PATH=path_to_your_python_bin_directory:${PATH} Radhikha

RadhikhaMyneni · ‎02-25-2020

That is correct, this capability tries different flow architectures with various model types and also optimizes for the best hyperparameters for each model type. It then picks the top 5 (5 by default but that number is configurable) models and draws those flows as a pipeline in Model Studio. Radhikha

RadhikhaMyneni · ‎02-21-2020

I would say RPM in EM is similar to the basic, intermediate and the advanced templates that are already shipped with Model Studio. What Jonathan is referring here is a new capability that was added in Model Studio as part of VDMML 8.5 release that implements the autoML (automated machine learning) initiative; where the pipeline is built after trying different combinations of data-preparation, modeling and hyperparameter turning steps for the input data specified. How many combinations it tries depends on how much time you give for the maxModelingTime parameter. Radhikha

RadhikhaMyneni · ‎12-18-2019

Question How can I pass data from an Open Source Code node to next node in SAS Model Studio? Answer Using the “Use output data in child nodes” property available in SAS Visual Data Mining and Machine Learning 8.5 or later. When this property is selected, user can write code in Python or R and should make the output data available in the dm_scoreddf data frame (if “Generate data frame” property is selected) or the node_scored.csv file. While executing, the node saves this output data and uses it as input in subsequent child node(s). Even though data can be passed through the Open Source Code node with this new property, capabilities such as download score code, download score API, register models, publish models or score holdout data will not be enabled for open source models as there is no underlying SAS score code for them. An example of passing data created in Open Source Code node in Data Mining Preprocessing lane to subsequent modeling node (Forest and Gradient Boosting) can be found on GitHub at sas-viya-dmml-pipelines project. Example on GitHub

RadhikhaMyneni · ‎04-24-2019

Question Is it possible to create indicator variables (for each input) for missing values without actually imputing in Model Studio? Answer Yes, add Imputation node to the pipeline, change the following properties and run the node. Under Class Inputs, change Default method to none Under Interval Inputs, change Default method to none Under Indicators, (1) select Unique indicators check-box, (2) set Indicator subject to Missing variables and (3) set Indicator role to Input. You will see that there is a new variable (with M_ prefix) for every input that has a missing value in the training data. This new variable takes 0 or 1 values to indicate missingness (0=not missing, 1=missing).

RadhikhaMyneni · ‎04-03-2019

Question Can the R ggplot package be used in the Open Source Code node for visualization purposes? Answer Yes, you can use the Open Source Code node in Model Studio to visualize plots using ggplot2 package in R software. Other Python or R visualization packages can also be used. The plots need to be saved as image files with rpt_ prefix and .png, .jpeg or .gif file extension. They will be displayed in the Results after the successful execution of the Open Source Code node. Multiple visualizations using ggplot2 and rpart.plot packages are shown in this GitHub example. Example on GitHub Note that the above example requires pre-installation of rpart, rpart.plot and ggplot2 packages in R software on the Compute Server. A special thanks to @MelodieRush for sharing similar example using SAS Enterprise Miner that led to this work.

RadhikhaMyneni · ‎03-20-2019

Question For an existing project, would it be possible to: Reload the input table with additional records? Reload the input table with additional variables? Load the input table with different name? Change the way data is partitioned after any pipeline is executed? Answer The answer to all the above four questions is YES! To change the data source or data partition for an existing project: Open the existing Project and go to the Data tab. Click on the “Data sources” icon on the left. Click on “Replace data source” icon on the top left – this gives you the capability to select a new data source; whether it is table with additional records, table with additional variables or table with a new name. Additionally, if you want change how the data is partitioned, follow the above steps and select the same or a new table. Then go to the “Partition Data” tab in Project settings (Top right icon to modify accordingly. Note that the target variable role in a Model Studio project cannot be modified after executing any node or pipeline in it.

Online Status	Offline
Date Last Visited	‎01-03-2023 07:15 PM

SAS Support Communities

Follow Us

What is...