NaNs during training when running the TinyYoloHouseNumberDetection example

[b005t3r]

Issue Description

Please describe your issue, along with:

• expected behavior: the network trains normally, no NaNs are reported
• encountered behavior: training crashes after the first fit() call because of NaNs (I enabled NaN detection, see below)

Version Information

Please indicate relevant versions, including, if relevant:

• Deeplearning4j version: 1.0.0-beta7 and 1.0.0-M2.1
• platform information: macOS Sonoma 14.2.1

The mentioned example:
https://github.com/deeplearning4j/deeplearning4j-examples/blob/master/dl4j-examples/src/main/java/org/deeplearning4j/examples/advanced/modelling/objectdetection/TinyYoloHouseNumberDetection.java

You'll need to add this as the first call in the main() method:

        Nd4j.getExecutioner().setProfilingConfig(ProfilerConfig.builder()
                .checkForINF(true)
                .checkForNAN(true)
                .checkElapsedTime(true)
                .checkLocality(true)
                .checkWorkspaces(true)
                .build());