When calling logPrediction using the nodejs library, I passed the option parameter in with computeFeatureContributions: true, however I forgot to add threshold too.
When I ran the code I got back a generic Error: bad request. It took me a little bit of fiddling with double checking urls, the app was running, etc before I checked the examples and tried adding threshold to see if it would help.
I initially assumed I wouldn't need the threshold because tuning was not supported for my model type in the first place.
Once I added threshold my code was working again.
Is there a better error that can be provided to clients if they are accidentally misusing the API?
When calling
logPrediction
using the nodejs library, I passed the option parameter in withcomputeFeatureContributions: true
, however I forgot to addthreshold
too.When I ran the code I got back a generic
Error: bad request
. It took me a little bit of fiddling with double checking urls, the app was running, etc before I checked the examples and tried adding threshold to see if it would help.I initially assumed I wouldn't need the threshold because tuning was not supported for my model type in the first place.
Once I added threshold my code was working again.
Is there a better error that can be provided to clients if they are accidentally misusing the API?