Problem Note 67296: An exception might occur when you score a Convolutional Autoencoder model where results might not return back to the client application
 |  |  |  |  |
When you run a dlScore action to score a Convolutional Neural Networks (CNN) model for an autoencoder segmentation task (where input and output variables are identical), an exception might occur on the CAS server. This issue occurs when the model is trained with data specifications (via the dataSpecs= option).
For example, you might encounter this issue when you run code similar to the following:
/*****************************/
/* Build a model shell. */
/*****************************/
BuildModel / modeltable={name='denoise_autoencoder', replace=1} type = 'CNN';
/***********************************/
/* Add an input layer. */
/***********************************/
AddLayer / model='denoise_autoencoder' name='input1' layer={type='input' nchannels=3 width=416 height=416 };
/************************/
/* Add hidden layers. */
/************************/
AddLayer / model='denoise_autoencoder' name='ConVLayer1' layer={type='CONVO' nFilters=4 width=5 height=5
stride=2 act='ELU' init='MSRA2' } srcLayers={'input1'};
...more layers here...
AddLayer / model='denoise_autoencoder' name='Convo7' layer={type='CONVO' nFilters=3
width=1 height=1 stride=1 act='Identity' includeBias=FALSE init='MSRA2'} srcLayers={'BNConvo6'};
/********************************************/
/* Add an Output layer (segmentation). */
/********************************************/
AddLayer / model='denoise_autoencoder' name='outlayer' layer={type='segmentation'} srcLayers={'Convo7'};
quit;
proc cas;
dlTrain / table={name='Cleaned_and_shuffled'} model='denoise_autoencoder'
modelWeights={name='ConVTrainedWeights_d', replace=1}
bestweights={name='Trainedautoencoder_w', replace=1}
dataSpecs={
{layer='outlayer', dataLayer='input1', type='IMAGE'},
{data={'_image_'}, layer='input1',type='IMAGE'}
}
optimizer={minibatchsize=5, loglevel=2,
algorithm={method='Adam', lrpolicy='Step', gamma=.7, stepsize=4, learningrate=.007},
maxepochs=1}
seed=12345;
quit;
proc cas;
dlScore / table={name='Cleaned_and_shuffled'}
model='denoise_autoencoder'
initWeights='Trainedautoencoder_w'
layerOut={name='OutputImages',caslib='imagelib'}
layers={'input1'}
layerImageType='BASE64'
casout={name='ScoredData', replace=1};
quit;
After the dlScore action runs, the results might never resurface to the client application. For example, if you click the Code tab to access the SAS® Studio code editor, a pop-up dialog box might be displayed after a while (for example, about 30 seconds), indicating that the connection to the server is lost.
SAS Studio
Error: Unable to connect to server. 502:Proxy Error
Click the Hot Fix tab in this note for a link to instructions about accessing and applying the software update.
Operating System and Release Information
| Product Family | Product | System | Product Release | SAS Release | ||
| Reported | Fixed* | Reported | Fixed* | |||
| SAS System | SAS Visual Data Mining and Machine Learning | Linux for x64 | 8.5 | 2020.1.3 | Viya | Viya |
| Microsoft® Windows® for x64 | 8.5 | 2020.1.3 | Viya | Viya | ||