The cross-entropy loss divided by 10 is not a loss function that can be used for machine learning ? True?
The cross-entropy loss MULTIPLIED by 10 is not a loss function that can be used for machine learning?
Whether we multiply or divide a loss function by scalar does not matter. Either way, the resulting function would be a valid loss function! What is important for a function to be a loss is that it is lower for better results, and higher for worse results.