We run PAL Multilayer Perceptron as an AMDP service. The training of the model fails when the parameter MOMENTUM_FACTOR is not 1.0. The program ends with a WP crash. The WP is then restarted. The problem does not occur when MOMENTUM_FACTOR = 1.0. Is this problem known?
In any case the whole process of training is very memory intensive on ABAP side. We are using a dataset of 100MB. A normal ABAP user context of 2 Gb is not enough. I had to enhance the user context to 8GB. Is there any explanation why this procedure needs so much memory on ABAP side although the procedure is running on DB side?
Thx. Thomas
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.