Finishes before thinking ends.
#2
by
visionarchitect - opened
Sorry for dumb question - I am new here
I have Android implementation that is answering first prompt correctly and in another prompt it breaks right at the end of thinking (it does not answer - end thinking tag not displaying but onDone is called
Can anyone help?
Is there a way to prevent it from doing so complex reasoning?
Do you know approximately how many tokens were inputted / generated by the model by the time it stops responding? It looks like the model should have a 4k context length. I wonder if you are hitting the end of the context length and that is why it stops early.