I’m running into an error repeatedly with the Keras Network Learner:
ERROR Keras Network Learner 3:18 View instantiation failed: Range(double, double): require lower (0.0) <= upper (-12.753908157348633).
In particular “View instantiation failed” which appears to be an error with visualizing the results of the learner. Has anyone else run into this?
Thanks for posting. Could you maybe specify in which cases you encounter that? Or maybe provide the workflow? Makes it easier for us to reproduce the error and looking into it
Thanks a lot!
I’m not able to provide a workflow; I ran into it multiple times while building a keras predictor into multiple classes. What’s really strange is that the epic count and batch size seemed to be related to the issue.
At which configuration for batch size, epochs etc. did that happen?
Can’t remember by now. Won’t have time to go back and test this out further any time soon, so the thread can just be deleted.
Thanks a lot anyway for reporting!
Managed to have a closer look and play around with it in the meanwhile, and found that the view throws an error if the loss is negative. That’s a bug and it’s reported to the development team, so that will be fixed
Yeah, it was almost certainly an error due to bad model configuration but I figured I’d bring it up in the forum anyways. Thank you for submitting the bug report!