Finetuning produces noisy output
I have finetuned mistral 7b model using domain based finetuning in amazon jumpstart. But it is producing nosiy output like this
{
"question": "What is maximum benefit?",
"answer": "$7,500"
},
{
"question": "What is Maximum Benefit Period duration with Age at Disablement and Duration in Benefits?",
"answer": " listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeivalue: listadeiv
What could be the reason and how to avoid it? Is there any parameters which we need to adjust?